JP2001312289A - Filter circuit for band division and signal analyzer and signal processing device using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal analyzer capable of performing high-resolution frequency analysis while suppressing the complexity of an arithmetic circuit, and to obtain a signal in which a specific change is given to an original signal by synthesizing the analyzed signal. A signal processing device is provided. SOLUTION: The following processing is performed to divide an input signal into a plurality of narrow bands and extract signals of each narrow band. First, the switch 220 determines the supply destination of the input signal to each of the delay units 231 to 23.
N is switched cyclically every sampling period T. Each signal after the delay by each of the delay units 231 to 23N passes through each of the filters 241 to 24N corresponding to each narrow band,
Further, after being multiplied by a predetermined constant by the multipliers 251 to 25N, it is input to the IDFT unit 260 which performs N discrete inverse Fourier transform. Of the 0th to N-1th signals output from the IDFT unit 260, the odd-numbered signals are output from the analysis filter bank 200 via the spectrum inverters 272 to 27N, and the even-numbered signals are output as they are. You.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a band dividing filter circuit for dividing a frequency band of an input signal into a plurality of narrow bands and outputting a signal corresponding to each narrow band component of the signal, and The present invention relates to a signal analyzer for performing frequency analysis of a signal using such a band dividing filter circuit and a signal processing device such as a graphic equalizer or an effector for giving a specific change to a signal.

[0002]

2. Description of the Related Art In recent years, with the advance of digital technology and higher performance of microprocessors, various kinds of signal processing are often realized by digital signal processing. When digital signal processing is performed, frequency analysis of the signal is often performed for the purpose of grasping in advance the properties of the signal to be processed.

In the field of digital signal processing, a discrete Fourier transform (hereinafter referred to as "DFT") is used for this frequency analysis.
And discrete inverse Fourier transform (hereinafter referred to as "IDFT")
Is often used. When a signal is subjected to frequency analysis using DFT and IDFT, there is a limit on the relationship between the signal to be analyzed and the wavelength, and the DFT and IDFT for the N-point samples can accurately analyze the frequency components of the signal. , N samples are only frequency components corresponding to integer wave numbers. In general, a signal to be analyzed contains frequency components having a non-integer wave number, so that these frequency components cannot be analyzed with high accuracy.
Therefore, a signal including a frequency component of a non-integer wave number is
When frequency analysis is performed using T and IDFT, a target signal is multiplied by a window function. As a result, it is possible to analyze even non-integer wave number frequency components with high accuracy. However, in this case, since the characteristics of the window function are superimposed on the result of the frequency analysis, there is a problem that a device for eliminating the influence of the characteristics of the window function is required to correctly grasp the result of the frequency analysis.

[0004] In addition to using DFT and IDFT, a band-pass filter having a very narrow band-pass characteristic
It is also possible to perform frequency analysis using (BPF). For example, a plurality of Bs respectively corresponding to the bands to be frequency-analyzed
A frequency analysis can be performed by preparing a PF and observing the outputs of the plurality of BPFs comprehensively. In this case, the above-described problem that has occurred when DFT and IDFT are used does not occur.

FIG. 22 is a block diagram showing a configuration of a conventional signal analyzer for performing frequency analysis using a plurality of BPFs. As shown in FIG. 22, the conventional signal analyzer has an input terminal 1100, BPFs 1101 to 11
0N, the amplitude detectors 1301 to 130N, and the output terminal 1
401 to 140N.

Hereinafter, the operation of the conventional signal analyzer will be described with reference to FIG. The signal to be analyzed is input terminal 1
100 BPFs 1201 to 120N
Supplied to. Each of the BPFs 1201 to 120N has a pass characteristic of passing only a frequency component in a different band. Is output as
The amplitude value of the frequency component extracted for each band in this manner is
The amplitude values detected by the amplitude detectors 1301 to 130N are output from the output terminals 1401 to 140N, respectively.
N respectively.

The above example is configured to analyze the amplitude. However, in the case of analyzing the phase, if the phase detector is used instead of the amplitude detector, the signal analysis can be performed in the same manner. It is.

As described above, by using a plurality of BPFs having different pass characteristics, it is possible to separate a signal into components of a plurality of bands, detect an amplitude and a phase of each band, and perform signal analysis. It becomes possible.

[0009]

In order to increase the frequency resolution in the above-described conventional signal analyzer, it is necessary to divide the signal into narrower bands. To split the signal into narrower bands, each BPF 1201-120N
Must be narrower. For example, 1
If a 000 Hz signal is divided into 100 Hz bands, ten BPFs with a 100 Hz pass band are required. However, if a higher frequency resolution is used to divide the signal into 10 Hz bands, a pass band is required. Is 10 Hz BPF is 1
00 pieces are required.

As described above, when a higher frequency resolution is required in the conventional signal analyzer, a large number of BPFs having a narrower pass band are required. In general, the narrower the passband BPF, the higher the order of the BPF. In other words, a larger amount of calculation is required, which results in a complicated and large-scale BPF circuit configuration. Therefore, there is a problem that the cost of the BPF increases.

[0011] As a method for avoiding the above problem,
By preparing only one BPF and changing its passing characteristic over time, it is also conceivable to perform a sweep-type signal analysis in which a calculation equivalent to a calculation of a plurality of BPFs is performed by only one BPF. However, in this case, since a large amount of arithmetic processing required for analyzing all bands is all processed by only one BPF, it takes time, and as a result, real-time signal analysis is impossible. There is a problem.

Therefore, the present invention can increase the frequency resolution while suppressing the complexity and the scale of the arithmetic circuit,
Further, it is an object of the present invention to provide a signal analyzer capable of real-time signal analysis. It is another object of the present invention to provide a signal processing device that gives a specific change to a signal by utilizing the technique of frequency analysis in such a signal analysis device.

[0013]

According to a first aspect of the present invention, a frequency band of an input signal sampled at a predetermined sampling period T is divided into a predetermined number N of narrow bands, and each narrow band in the input signal is divided. A band dividing filter circuit for generating N narrow-band signals each corresponding to a band component, wherein each of the supplied signals is delayed by a predetermined delay amount from a 0th to N-1.
Nth delay units and the delay unit that receives the input signal and supplies it to one delay unit selected from the 0th to N-1th delay units, and the selected delay unit For cyclically switching between the 0th to N-1th delay units for each sample value of the input signal, and a delay output from the 0th to N-1th delay units N-th from the 0th through which the subsequent signals pass
The signals after passing through the first N digital filters and the 0th to N-1th digital filters are respectively multiplied by predetermined constants. Perform N discrete inverse Fourier transforms on the multiplied 0th to N-1th signals output from the N multipliers and the 0th to N-1th multipliers, respectively. And a discrete inverse Fourier transformer, wherein N signals from 0th to N-1th obtained by the discrete inverse Fourier transform are output as N narrowband signals.

According to the first aspect, the input signal is decimated by the switch to obtain N signals whose sampling frequency has been reduced to 1 / N. On the other hand, predetermined processing is performed by each delay unit, each digital filter, and each multiplier, and then N discrete inverse Fourier transforms are performed on N signals obtained by the processing, and the resulting signals are processed. As a result, a signal corresponding to each narrow-band component in the input signal is obtained. That is, when the frequency band of the input signal is equally divided into N narrow bands, the sampling theorem allows each narrow band to have a sampling frequency of 1 without losing information.
/ N, the present invention generates each narrow-band signal by performing the above-described processing on N signals obtained by thinning out the input signal. In this manner, a process equivalent to simultaneous execution of extraction of each narrowband signal by the BPF and thinning of the sampling is performed, so that an arithmetic circuit for realizing a filter circuit for band division becomes complicated and large. It is possible to increase the frequency resolution while suppressing the scale, and, unlike the sweep method described above, can extract a signal component corresponding to each narrow band in real time.

According to a second aspect of the present invention, in the first aspect, the odd number of the 0th to (N-1) th N signals after the discrete inverse Fourier transform output from the discrete inverse Fourier transformer is used. "-1" every 1 sample for the signal
And an even-numbered signal output from the discrete inverse Fourier transformer and an odd-numbered signal after being multiplied by “−1” every other sample by the spectrum inverter. N signals consisting of
It is characterized by outputting as N narrow-band signals.

According to the second aspect of the present invention, the N signals consisting of the even-numbered signals output from the discrete inverse Fourier transformer and the odd-numbered signals whose spectrum has been inverted by the spectrum inverter are obtained. Since the signals are output as N narrow-band signals, an accurate narrow-band signal can be obtained.

In a third aspect based on the first or second aspect, the bandwidths of the N narrow bands are all equal, and the delay amount of the i-th one of the N delay units is Assuming that the delay amount is expressed using a unit delay operator z ^-1 corresponding to the sampling period T, z ⁻⁰ when i = 0 and z ^{− (} when 1 ≦ i ≦ N−1. ^Ni)
, And the said bandwidth of each narrowband and f _B, -f _B / 2
The characteristics of the digital filter from the frequency of the bandpass to passband a frequency up to f _B / 2

(Equation 7) And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
If the characteristic represented by H _i (z) = H (−jz ^N ) _i and i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. The signal that has passed through the digital filter is multiplied by e ^{−j [} iπ ^{/ ( 2N)]} as the constant.

According to the third aspect, the bandwidths of the N narrowband signals are all the same, and the characteristics of each digital filter for obtaining each narrowband signal are the basic bandpass digital filter. It is obtained by frequency-shifting the characteristic H (z) of the filter (called a prototype filter). Thereby, a plurality of Bs used in the conventional band division are
Since PF is replaced by a plurality of filter operations corresponding to at most one BPF as an order and an inverse Fourier transform,
The amount of calculation for realizing the band dividing filter circuit is greatly reduced.

In a fourth aspect based on the first or second aspect, the bandwidths of the N narrow bands are all equal, and the delay amount of the i-th one of the N delay units is Assuming that the delay amount is expressed using a unit delay operator z ^-1 corresponding to the sampling period T, z ⁻⁰ when i = 0 and z ^{− (} when 1 ≦ i ≦ N−1. ^Ni)
, And the said bandwidth of each narrowband and f _B, -f _B / 2
The characteristics of the digital filter from the frequency of the bandpass to passband a frequency up to f _B / 2

(Equation 8) And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is 0
Assuming that the given integer satisfies ≦ i ≦ N−1, the i-th multiplier multiplies the signal passed through the i-th digital filter by e ^{−j [i} π ^{/ N ]} as the constant. It is characterized by.

According to the fourth aspect of the invention, the bandwidths of the N narrowband signals are all the same, and the characteristics of each digital filter for obtaining each narrowband signal are based on the basic bandpass digital filter. It is obtained by frequency-shifting the characteristic H (z) of the filter (called a prototype filter). Thereby, a plurality of Bs used in the conventional band division are
Since PF is replaced by a plurality of filter operations corresponding to at most one BPF as an order and an inverse Fourier transform,
The amount of calculation for realizing the band dividing filter circuit is greatly reduced.

According to a fifth aspect of the present invention, a predetermined sampling period T
A signal analyzer for performing frequency analysis on an input signal sampled at
The band dividing filter circuit according to any one of the third to third aspects, and a root mean square value of a real part and an imaginary part of each value of the N narrow band signals output from the band dividing filter circuit. , And N number of mean squares output by

According to the fifth aspect, the mean square value of the real part and the imaginary part of the value of each narrow band signal output from the band dividing filter circuit is output from the root mean squarer.
If each narrowband signal is an analysis signal, the amplitude of each narrowband signal can be continuously detected with a small amount of calculation per unit time. In the case where the band splitting filter circuit of the third invention is used in the signal analyzing apparatus of the fourth invention, the i-th narrow band (0
≤ i ≤ N-1) The transfer function H _i (z) of the digital filter for extracting
_k (z) from the not have a pass band in the band 0~-f _B / 2, each narrow band signal output from the band-splitting filter circuit is analytic signal. In this case, each narrowband signal has a real number component and an imaginary number component, and the amplitude can be calculated as a half of the sum of squares.

According to a sixth aspect of the present invention, a predetermined sampling period T
A signal analyzer for performing frequency analysis on an input signal sampled at
And a ratio I / I between a real part R and an imaginary part I of each of the N narrowband signals output from the band dividing filter circuit. And N phase detectors each outputting a value indicating an arctangent of R.

According to the sixth aspect, the inverse tangent of the ratio I / R between the real part R and the imaginary part I of the value of each narrow band signal output from the band dividing filter circuit is output from the phase detector. Therefore, if each narrow-band signal is an analysis signal, the phase of each narrow-band signal can be detected continuously with a small amount of computation per unit time. When the band splitting filter circuit of the third aspect is used in the signal analyzing apparatus of the fifth aspect, each narrow band signal output from the band splitting filter circuit is similar to the fourth aspect. Is an analytic signal, each narrowband signal has a real number component and an imaginary number component, and the phase can be calculated as the inverse tangent of the ratio.

According to a seventh aspect, a predetermined sampling period T
A signal analysis device for performing frequency analysis on the input signal sampled at
And the N number of N-band signals each outputting the root-mean-square value of the real part and the imaginary part of each of the N narrow-band signals output from the band-division filter circuit. A mean squarer.

According to the seventh aspect, the mean square value of the real part and the imaginary part of the value of each narrow band signal output from the band dividing filter circuit is output from the root mean squarer.
If each narrowband signal is an analysis signal, the amplitude of each narrowband signal can be continuously detected with a small amount of calculation per unit time. In the case where the band dividing filter circuit of the fourth invention is used in the signal analyzing apparatus of the seventh invention, the i-th narrow band (0
≤ i ≤ N-1) The transfer function H _i (z) of the digital filter for extracting
_k (z) from the not have a pass band in the band 0~-f _B / 2, each narrow band signal output from the band-splitting filter circuit is analytic signal. In this case, each narrowband signal has a real number component and an imaginary number component, and the amplitude can be calculated as a half of the sum of squares.

According to an eighth aspect of the present invention, a predetermined sampling period T
A signal analysis device for performing frequency analysis on the input signal sampled at
And the inverse tangent of the ratio I / R between the real part R and the imaginary part I of each value of the N narrow band signals output from the band dividing filter circuit of the present invention. N phase detectors each outputting a value.

According to the eighth aspect, the arctangent of the ratio I / R of the real part R and the imaginary part I of the value of each narrow band signal output from the band dividing filter circuit is output from the phase detector. Therefore, if each narrow-band signal is an analysis signal, the phase of each narrow-band signal can be detected continuously with a small amount of computation per unit time. When the band splitting filter circuit of the fourth aspect is used in the signal analyzing apparatus of the eighth aspect, each narrow band signal output from the band splitting filter circuit is similar to the seventh aspect. Is an analytic signal, each narrowband signal has a real number component and an imaginary number component, and the phase can be calculated as the inverse tangent of the ratio.

In a ninth aspect of the present invention, a predetermined sampling period T
A signal processing apparatus for generating a modified signal of the input signal by giving a specific change to the input signal sampled in the above, wherein the frequency band of the input signal is equally divided into a predetermined number N of narrow bands, 0th to N-1st Ns corresponding to each narrowband component in the signal
A band dividing filter circuit for generating N narrowband signals from the input signal, the narrowband signals having N times the sampling period T of the input signal; And N output points from 0th to N-1th, and the 0th to N-1th narrowband signals are 0th to N-1th.
A mapper for receiving the narrowband signal received at each of the input points up to the first and outputting the narrowband signal received from each output point at an input point corresponding to each output point by a predetermined mapping; and N-th
A signal for generating a modified signal of the input signal is generated by assuming that signals output from the first output point correspond to the 0th to N−1th narrowband signals, respectively. And a filter circuit.

According to the band dividing filter circuit of the ninth aspect, each narrow band component in the input signal is
It is generated as N narrowband signals having a sampling period N times the sampling period T of the input signal. That is, a signal expressing each narrow band component in the input signal in a signal base band (frequency band from a frequency of 0 to a frequency corresponding to one narrow band width) is generated as each narrow band signal having the same bandwidth. You. As a result, various frequency components of the input signal can be assigned to frequencies different from the original frequency (the input signal can be converted in the frequency domain). Therefore, in the ninth aspect, mapping is performed between N outputs of the band dividing filter circuit for generating each narrow band signal and N inputs of the synthesizing filter circuit. By combining the band signals, a modified signal of the input signal is generated. Therefore, unlike a conventional signal processing device in which a signal is processed by conversion on the time axis, the signal is processed by conversion in the frequency domain, and the input signal is not compressed or expanded in the time axis direction. Therefore, it is possible to process the input signal more flexibly (processing such as frequency conversion and graphic equalizer) than before.

In a tenth aspect based on the ninth aspect, the mapper is a function g (y) in which a predetermined set including the integer y indicating the output point is defined as a domain and a predetermined set of real numbers is defined as a range. )), The signal to be output from the j-th output point is N-th from the 0-th output point.
Generating by interpolation based on the real number g (j) and the plurality of narrowband signals received at the plurality of input points corresponding to integers near the real number g (j) among the first input points; It is characterized by.

According to an eleventh aspect, in the tenth aspect,
The mapper generates a signal MO _j to be output from the j-th output point by linear interpolation represented by the following equation: MO _j = MI _i + α · (MI _{i + 1} −MI _i ) i = [g (j)] α = {g (j) − [g (j)]} where j is an integer satisfying 0 ≦ j ≦ N−1, and x
= G (y) is a function defining the mapping in the mapping unit, and a domain including all integers y satisfying 0 ≦ y ≦ N−1 is defined as a domain and only real numbers x satisfying 0 ≦ x <N−1 [G (j)] is a function whose range is a set including
And MI _i is the _ith narrowband signal received at the ith input point in the mapper.

According to the tenth or eleventh aspect,
A signal output from each output point of the mapper is generated by interpolation using a plurality of narrowband signals received at a plurality of input points associated with the output point by a predetermined mapping, so that the input signal is Conversion in the frequency domain for processing can be performed more accurately, and as a result, the quality of the output signal from the signal processing device is improved.

In a twelfth aspect based on the ninth aspect, in the ninth aspect, the mapping device receives time information from the outside, and the correspondence between the input point and the output point given by the mapping in the mapping device is: It changes with the passage of time based on time information.

According to the twelfth aspect, as in the ninth aspect, various frequency components of the input signal are allocated to different frequencies from the original frequency and synthesized to generate a deformed signal and The mapping by the mapper between the N outputs of the band dividing filter circuit and the N inputs of the combining filter circuit changes over time. For this reason, the above-mentioned frequency assignment changes with time. That is, the frequency component of the input signal is dynamically converted without being compressed or expanded in the time axis direction, thereby generating a modified signal of the input signal. Such dynamic conversion of frequency components is very effective as an effect for electronic devices and the like.

According to a thirteenth aspect, in the twelfth aspect,
The mapper is a function g (y, t) having a predetermined set of pairs of an integer y indicating the output point and a value t indicated by the time information as a domain and a predetermined set of real numbers as a range. Let the signal to be output from the j-th output point be N-1 from the 0-th.
By interpolation based on the plurality of narrowband signals received at a plurality of input points corresponding to integers near the real number g (j, t) and the real number g (j, t) among the input points up to the th, It is characterized by generating.

According to a fourteenth aspect, in the thirteenth aspect,
The mapper generates a signal MO _j to be output from the j-th output point by linear interpolation represented by the following equation: MO _j = MI _i + α · (MI _{i + 1} −MI _i ) i = [g (j, t)] α = {g (j, t) − [g (j, t)]} where j is an integer satisfying 0 ≦ j ≦ N−1. Yes, x
= G (y, t) is a function that defines the mapping in the mapping device, and is a set including all pairs of integers y satisfying 0 ≦ y ≦ N−1 and values t indicated by the time information. [G (j, t)] is a function whose range is a set including only real numbers x satisfying 0 ≦ x <N−1 as a domain, and g (j, t) is g (j,
is the largest integer not exceeding t), and MI _i is the _ith narrowband signal received at the ith input point in the mapper.

According to the thirteenth or fourteenth aspect,
A signal output from each output point of the mapper is generated by interpolation using a plurality of narrowband signals received at a plurality of input points associated with the output point by a predetermined mapping, so that the input signal is Conversion in the frequency domain for processing can be performed more accurately, and as a result, the quality of the output signal from the signal processing device is improved.

A fifteenth invention is a signal processing apparatus for generating a modified signal of an input signal by giving a specific change to the input signal sampled at a predetermined sampling period T, wherein the frequency band of the input signal is Is a predetermined number N
And N equal to the N-1th narrowband signal corresponding to each narrowband component of the input signal, and N times the sampling period T of the input signal. A band dividing filter circuit for generating N narrowband signals having the sampling period from the input signal, and amplitude adjustment information received from the outside, and based on the amplitude adjustment information, the 0th to N-1th The amplitude of the narrow-band signal is adjusted, and N-
By synthesizing an amplitude adjustment circuit that outputs the first narrowband signal and the 0th to N−1th narrowband signals output from the amplitude adjuster, a modified signal of the input signal is obtained. And a synthesizing filter circuit.

According to a sixteenth aspect, in the fifteenth aspect,
The amplitude adjusting circuit includes N number of 0-th to (N-1) -th amplitude adjusters for adjusting the amplitudes of the 0th to N-1th narrow-band signals output from the band dividing filter circuit. And the 0th to N-1st N corresponding to the 0th to N-1th narrowband signals output from the band dividing filter circuit, respectively.
Are received as the amplitude adjustment information, and i is set to 0
If an arbitrary integer satisfying ≦ i ≦ N−1, the ith amplitude adjuster adjusts the amplitude of the ith narrowband signal based on the ith adjustment value.

According to the fifteenth or sixteenth aspect,
The amplitude of each narrow band signal obtained from the input signal is adjusted based on the amplitude adjustment information, and the adjusted narrow band signals are combined. That is, the input signals are combined after their amplitudes of various frequency components are changed from their original values, thereby generating a modified signal of the input signal. Thus, by changing the amplitude of the input signal for each narrow band, a function as a so-called graphic equalizer can be realized.

In a seventeenth aspect based on the ninth or fifteenth aspect, the band dividing filter circuit is configured to delay the signals supplied thereto by a predetermined delay amount from the 0th to the (N-1) th. N delayers up to
The input signal is received and supplied to one delay device selected from the 0th to N-1th delay devices, and the selected delay device is selected from the 0th to N-1th delay devices. A switching unit that cyclically switches between the delay units for each sample value of the input signal, and a 0th to N−1th signal through which signals output from the 0th to N−1th delay units respectively pass The signals after passing through the N digital filters up to the Nth and the digital filters from the 0th to the N-1th are respectively multiplied by predetermined constants. N-point discrete inverse Fourier transform is performed on the multiplied 0th to N-1th signals output from the 0th to N-1th multipliers, The discrete inverse Fourier 0th a signal 換後 N-
A discrete inverse Fourier transformer for outputting N signals up to the first, and N output from the discrete inverse Fourier transformer
And a spectrum inverter that multiplies the odd-numbered signal of the number of signals by "-1" every other sample,
N consisting of an even-numbered signal output from the discrete inverse Fourier transformer and an odd-numbered signal after being multiplied by “−1” every other sample by the spectrum inverter.
Signals are output as the N narrowband signals.

According to an eighteenth aspect, in the seventeenth aspect,
Assuming that the delay amount of the i-th delay unit among the N delay units is expressed by using a unit delay operator z ⁻¹ corresponding to the sampling period T, i =
When 0 is a z ^-0, when 1 ≦ i ≦ N-1 is z ^- a ^(Ni), the bandwidth of each narrowband and f _B, -
The characteristics of the digital filter of the band-pass type that the frequency of f _B / 2 and pass band up to a frequency of f _B / 2

(Equation 9) And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
If the characteristic represented by H _i (z) = H (−jz ^N ) _i and i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. The signal that has passed through the digital filter is multiplied by e ^{−j [} iπ ^{/ ( 2N)]} as the constant.

According to a nineteenth aspect, in the seventeenth aspect,
Assuming that the delay amount of the i-th delay unit among the N delay units is expressed by using a unit delay operator z ⁻¹ corresponding to the sampling period T, i =
When 0 is a z ^-0, when 1 ≦ i ≦ N-1 is z ^- a ^(Ni), the bandwidth of each narrowband and f _B, -
The characteristics of the digital filter of the band-pass type that the frequency of f _B / 2 and pass band up to a frequency of f _B / 2

(Equation 10) And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is 0
Assuming that the given integer satisfies ≦ i ≦ N−1, the i-th multiplier multiplies the signal passed through the i-th digital filter by e ^{−j [i} π ^{/ N ]} as the constant. It is characterized by.

In a twentieth aspect based on the ninth or fifteenth aspect, the synthesizing filter circuit outputs “−1” to the signal output from the odd-numbered output point of the mapper every one sample. For N signals consisting of a spectrum inverter to be multiplied, a signal output from an even-numbered output point in the mapping unit, and a signal multiplied by “−1” every one sample by the spectrum inverter, A discrete inverse Fourier transformer that performs N discrete inverse Fourier transforms and outputs N signals from 0th to N−1th, which are the signals after the discrete inverse Fourier transform; The 0th to N-1th signals output from the Fourier transformer are multiplied by predetermined constants, respectively.
Multipliers and 0th to N-1th N signals through which the 0th to N-1th signals after multiplication output from the 0th to N-1th multipliers respectively pass N digital delays 0 to N-1 for respectively delaying the signals after passing through each of the digital filters and the digital filters 0 to N-1 by a predetermined delay amount And a delayed signal output from one of the delay units selected from the 0th to N-1th delay units as a modified signal of the input signal, and A switching unit for cyclically switching the delay unit among the 0th to N-1th delay units for each sample value of the input signal.

According to a twenty-first aspect, in the twentieth aspect,
The delay amount due to the i-th delay unit out of the N delay units is expressed as follows: 0 ≦≦, where the delay amount is expressed using a unit delay operator z ⁻¹ corresponding to the sampling period T.
z ^-i for any integer i that satisfies i ≦ N-1,
The bandwidth of each narrowband and f _B, the -f _B / 2 of the characteristic of the digital filter bandpass of the frequency up to f _B / 2 and pass band from a frequency

[Equation 11] And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
If the characteristic represented by H _i (z) = H (−jz ^N ) _i and i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. The signal that has passed through the digital filter is multiplied by e ^{−j [} iπ ^{/ ( 2N)]} as the constant.

According to a twenty-second invention, in the twentieth invention,
The delay amount due to the i-th delay unit out of the N delay units is expressed as follows: 0 ≦≦, where the delay amount is expressed using a unit delay operator z ⁻¹ corresponding to the sampling period T.
z ^-i for any integer i that satisfies i ≦ N-1,
The bandwidth of each narrowband and f _B, the -f _B / 2 of the characteristic of the digital filter bandpass of the frequency up to f _B / 2 and pass band from a frequency

(Equation 12) And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is 0
Assuming that the given integer satisfies ≦ i ≦ N−1, the i-th multiplier multiplies the signal passed through the i-th digital filter by e ^{−j [i} π ^{/ N ]} as the constant. It is characterized by.

[0048]

Embodiments of the present invention will be described below with reference to the drawings. First, the first to fifth embodiments will be described with reference to FIGS. 1 to 15 (note that after the first to fifth embodiments, the filter characteristics are different from the first to fifth embodiments). The sixth to ninth embodiments having different conditions such as will be described.) (First Embodiment) FIG. 1 is a block diagram showing a configuration of a signal analyzer according to a first embodiment of the present invention. FIG.
As shown in FIG.
, An analysis filter bank 200, and a root mean square 311
To 31N and output terminals 401 to 40N.

The operation of the signal analyzer according to the first embodiment of the present invention will be described with reference to FIG. The signal to be analyzed is input via an input terminal 110 to the filter bank 20 for analysis.
Input to 0. The analysis filter bank 200 is a filter circuit for band division, divides a frequency band of an input signal into N narrow bands, and extracts a signal for each narrow band from the input signal. Mean squares 311 to 31N
Respectively detect the amplitude of the signal extracted for each narrow band. The amplitude of each detected narrowband signal is
The signals are output from the signal analyzer through output terminals 401 to 40N, respectively.

Hereinafter, the analysis filter bank 200 will be described in more detail. In the following, it is assumed that a signal input to the analysis filter bank 200 as a signal to be analyzed is a signal sampled at a sampling period T. Further, it is assumed that all N narrow-bandwidths f _B obtained by dividing the frequency band of the input signal to be analyzed are equal. The sampling frequency f _s = 1 / T of the bandwidth f _B and the input signal, it is necessary to satisfy the Nf _B ≦ f _s / 2 the sampling theorem. In the present embodiment, it is assumed to satisfy the following relationship with the sampling frequency f _s of the narrowband bandwidth f _B and the input signal. Nf _B = f _s / 2 These assumptions are the same in other embodiments described later.

FIG. 2 is a block diagram showing the configuration of the analysis filter bank 200 according to the first embodiment of the present invention. As shown in FIG. 2, this analysis filter bank 2
00 is an input terminal 210, a switch 220, and a delay 2
31 to 23N, filters 241 to 24N, and multiplier 2
51-25N, a discrete inverse Fourier transformer 260, a spectrum inverter 272-27N, and output terminals 281-2
8N.

When the delay units 231 to 23N are represented by D _i (i = 0, 1,..., N−1) using i as a variable, the delay amount in the delay unit D ₀ is z ⁻⁰ , Delay device D _i
The delay amount at (i ≠ 0) is z ^{− (Ni)} . Further, the filter 241～24N, i as variable respectively H _i
(I = 0, 1,..., N−1), the pass characteristic of the filter _Hi , that is, the transfer function H _i (z) is expressed by the frequency −f _B /
The characteristic, ie, transfer function, of a filter (hereinafter, referred to as a “prototype filter”) through which only signals from 2 to f _B / 2 pass

(Equation 13) Where H _i (z) = H (−jz ^N ) _i . Also, a multiplier _{251~25N, M i (i = 0,1} , ..., N-1) i as variable respectively expressed, the multiplier M _i
Is a constant E _i ^{-j (} iπ ^{/ 2N)} .

Next, the operation of the analysis filter bank 200 according to the first embodiment of the present invention will be described with reference to FIG.

The signal input to the analysis filter bank 200 is supplied to the switch 220 via the input terminal 210. The switch 220 sets the output destination of the supplied signal to the delay units 231, 232,..., 23N, 231, 232,.
, In a cyclic manner every time T. That is, a signal sampled at the sampling period NT is input to each of the delay units 231 to 23N. Delay device 2
The signals input to 31 to 23N are respectively supplied to filters 241 to 24N after being delayed by a predetermined time. Here, the delay amount z ^-1 is a time corresponding to the sampling period T of the original signal. The signals supplied to the filters 241 to 24N are respectively subjected to a predetermined filtering process, and then are multiplied by predetermined constants in the multipliers 251 to 25N. The signals multiplied by the constants in the multipliers 251 to 25N are N-point discrete inverse Fourier transformers (hereinafter, “IDF”).
(Referred to as a "T unit") 260 is subjected to discrete inverse Fourier transform.
An odd-numbered output of the 0-th to (N-1) -th outputs of the IDFT unit 260 is provided with a spectrum inverter 2
At 72 to 27N, -1 is multiplied every one sample. The even-numbered outputs of the IDFT unit 260 and the outputs of the spectrum inverters 272 to 27N are output from the analysis filter bank 200 via output terminals 281 to 28N corresponding to the respective outputs.

Hereinafter, the principle of frequency analysis in the signal analyzer according to the present embodiment, that is, the fact that the signal can be analyzed by the above-described operation of the signal analyzer will be described with reference to the drawings and mathematical expressions.

[0056] Generally, sampled signal at the sampling frequency f _s is the N-number of BPF, when output is equally divided for each band of 1 / N of the entire frequency band of the signal, the output signal Have only 1 / N the bandwidth of the original signal. Therefore, N B
When focusing on each of the PFs, the sampling frequency can be reduced to f _s / N by the sampling theorem without impairing the information amount of the output signal of each BPF. This operation of lowering the sampling frequency is thinning. The principle of this thinning will be described with reference to FIGS. In the following,
The N narrow bands obtained by equally dividing the entire frequency band of the input signal to the analysis filter bank 200 are divided into bands # 0, band # 1, band # 2,.
(N-1) (the same applies to other embodiments). As shown in FIGS. 3 and 4, the band extracted by the 0th BPF is band # 0, the band extracted by the first BPF is band # 2, and the band extracted by the second BPF is band # 0. The band extracted by the third BPF corresponds to band # 3, and corresponds to band # 1.

FIG. 3 is a diagram showing an example of the frequency spectrum of each narrow-band signal output from the BPF. FIG.
FIG. 4 is a diagram showing an example of a frequency spectrum after thinning out each narrowband signal output from the BPF. Sampled signal at the sampling frequency f _s is four BP
F (hereinafter referred to as “0th to 3rd BPFs”), when the signal is equally divided for each quarter band of the entire frequency band and output (that is, the entire frequency band has four narrow bands). ), The frequency spectrum of the output narrowband signal is the spectrum shown in FIG. Furthermore, the frequency spectrum of each narrow-band signal when the sampling frequency is reduced to f _s / 4, that is, when thinning is performed on each of the narrow-band signals output in this way, is the spectrum shown in FIG. .

[0058] A comparison of FIGS. 3 and 4, the spectrum of the signal output from the BPF in the case of FIG. 3, is stored in the signal base band (frequency 0 to F _B) in the case of FIG. 4 You can see that it is. However, the signal of band # 1 output from the third BPF and the second BPF
Regarding the signals in the band # 3 output from the PF, the spectrum in the signal base band after thinning out those signals is inverted. Therefore, by further inverting the spectrum, the spectrum before thinning is reduced to the signal base band. Is stored in As is well known, the signal spectrum can be inverted by multiplying the signal by -1 every other sample. As described above, even if the signals output from the BPF are thinned out, the spectra of the signals, that is, the properties of the signals are stored in the signal base band. Note that the spectrum inversion processing is performed in the spectrum inverters 272 to 27N.

In the above description, a signal having a spectrum as shown in FIG. 4 is extracted by thinning out the signal output from the BPF. Here, as described above, performing the thinning means changing the sampling frequency from f _s to f _s /
Reducing to N, that is, taking every Nth sample. Therefore, the remaining (N-1) samples are unnecessary. The amount of calculation for calculating the unnecessary samples increases as N increases, and as a result, the operation circuit becomes complicated and the cost increases. Therefore, in the present embodiment, of a plurality of operations corresponding to a series of operations of thinning out the signal output from the BPF, only the minimum necessary operation is performed by one filter bank, so that the operation circuit We are trying to simplify it.

Hereinafter, processing corresponding to a series of operations of extracting each narrow-band signal by the BPF and then thinning out the signal will be described below.
A description will be given, using mathematical expressions, that the use of the analysis filter bank 200 in the first embodiment makes it possible to perform only the minimum necessary calculations.

First, the input signal F (z) is

[Equation 14] And the transfer function H _k (z) of the _kth BPF is

(Equation 15) And the output signal T _k (z) from the k-th BPF is

(Equation 16) It expresses.

Equation (1) generally indicates that the input signal F (z) is
This shows that the samples constituting the input signal F (z) can be expressed as the sum of N sample sequences (0 sequence to N-1 sequence) taken out every Nth sample. The input signal F (z) is input in the order of..., N−2, N−1, 0, 1, 2,. Equation (2) indicates that the transfer function H _k (z) of the K-th BPF is generally expressed by N partial transfer functions (0 th to N th order) obtained by taking out operations of a plurality of orders in the BPF at every Nth order. (-1st order partial transfer function). In general, k <N
/ 2, the band extracted by the k-th BPF corresponds to band # 2k, and when k ≧ N / 2, the k-th BPF
Corresponds to band # [N-1-2 (k-N / 2)]. In addition, in the analysis filter bank 200, based on such a correspondence, the arrangement order of the output terminals of the signals extracted for each narrow band is band # 0, band # 1, band # 2,
.., And a circuit corresponding to each BPF and each output terminal are connected in the order of band # (N−1). Equation (3) generally states that the output signal T _k (z) from the k-th BPF is the output signal T _k (z) from the k-th BPF.
Can be expressed as the sum of N sample sequences (0 sequence to N-1 sequence) taken out every Nth sample. Note that the sequence of the input signal F (z) corresponds to the sequence of the output signal T _k (z) from the k-th BPF. For example, a signal output when a 0-sequence signal is input is 0 series.

The relationship between F (z), H _k (z) and T _k (z) is: T _k (z) = H _k (z) F (z)

[Equation 17] Becomes When this is expressed as a matrix,

(Equation 18) Becomes

In the equation (6), the output signal T _k (z)
Is a signal composed of N sample sequences ranging from 0 sequence to (N-1) sequence. According to the above-described sampling theorem, the properties of the signal can be preserved only by taking out one sample every N samples. That is, among the output signals in the equation (6),
It suffices that only T _k (z ^N ) ₀ is obtained by thinning. In this way, for each narrow band, only the output signal after being decimated is taken out, and a matrix expression is newly made for all the narrow bands.

[Equation 19] Becomes The matrix relating to H in the equation (7) represents a transfer function of a filter bank that performs processing by reducing the amount of calculation in consideration of thinning.

Here, paying attention to the fact that the digital filter H _k (z) for extracting each narrow band signal can be obtained by shifting the frequency of the prototype digital filter H (z) shown in FIG. Perform frequency conversion.

[0066] This conversion is the case of the k-th _{filter, f → f- (2f B ·} k + f B / 2) ... (8) becomes equivalent to the conversion. Here, z = e ^j ω ^T , ω = 2πf, T = 1 / (2Nf _B ) (9) Therefore, each of z and z ^N is z → z · W ^k · e ^j π ^{/ 2N} is converted ^{W = e -j2 π / N ...} (10) z N → -jz N ... (11).

Here, the transfer function H (z) of the prototype digital filter is

(Equation 20) And the matrix for the filter bank in equation (7) is:
When frequency conversion is performed,

(Equation 21) Becomes Here, the matrix [W] shown in Expression (14) represents a calculation process corresponding to one IDFT of N points.

From the above results, the arithmetic expression taking the thinning into account is:

(Equation 22) Becomes

As described above, the process of extracting a signal for each narrow band and then decimating the signal is performed by calculating the matrix [H], calculating the matrix [E], and calculating the matrix [W] (N-point IDFT).
Can be replaced with a series of processes combining.

Next, the operation amount of the matrix [H] will be considered. For example, if the number of band divisions N is 4 and the transfer function H (z) shown in equation (12) is realized by a 12th-order FIR (non-feedback type) filter, H (z) becomes H (z) ^{_{= a 0 + a 1 z -1}} + a 2 z -2 + a 3 z -3 + a 4 z -4 + a 5 z -5 + a 6 z -6 + a 7 z -7 + a 8 z -8 + a 9 z -9 + a 10 _{^{z -10 + a 11 z -11 +}} a 12 z -12 = (a 0 + a 4 z -4 + a 8 z -8 + a 12 z -12) + (a 1 z -1 + a 5 z -5 + a 9 z -9 ) + (A ₂ z ⁻² + a ₆ z ⁻⁶ + a ₁₀ z ⁻¹⁰ ) + (a ₃ z ⁻³ + a ₇ z ⁻⁷ + a ₁₁ z ⁻¹¹ ), so that H (z ⁴ ) ₀ = a ^{_{0 + a 4 z -4 + a}} 8 z -8 + a 12 z -12 H (z 4) 1 z -1 = a 1 z -1 + a 5 z -5 + a 9 z -9 H (z 4) 2 z -2 = a _{^{a 2 z -2 + a 6 z}} -6 + a 10 z -10 H (z 4) 3 z -3 = a 3 z -3 + a 7 z -7 + a 11 z -11. From this, when H (z) is an FIR filter, the amount of calculation of the matrix [H] is at most one filter H (z).
Therefore, according to the analysis filter bank 200 shown in FIG. 2, the calculation amount is greatly reduced, and the effect is particularly enhanced as the number N of band divisions increases.

As described above, the analysis filter bank 200 shown in FIG. 2 operates, and performs processing equivalent to simultaneous extraction of narrowband signals by the BPF and sampling thinning. As described above, in the analysis filter bank 200, the arrangement order of the output terminals of the signals extracted for each narrow band is band # 0, band # 1, band # 2, ..., band # (N-1). ), The circuits corresponding to the respective BPFs and the respective output terminals are connected.

Next, the operation of the first embodiment shown in FIG. 1 after the narrow-band signals are obtained by the analysis filter bank 200 will be described in more detail. k
The transfer function H _k (z) of the BPF is
(z) is derived by frequency conversion, so that band 0 to
It does not have the pass band to -f _s / 2. Thus, the output processed by H _k (z) has no component band 0~-f _s / 2, the so-called analysis signal (Analitic Signa
1). The analytic signal has a real number component and an imaginary number component, and the mean square of those components represents the amplitude. Therefore, by calculating the mean square of the real and imaginary components of the output signal of the analysis filter bank 200 in the root mean squares 311 to 31N, the amplitude of each narrow band signal can be obtained.

As described above, according to the first embodiment,
Processing equivalent to omitting unnecessary arithmetic processing can be realized, and a well-known fast Fourier transform algorithm can be applied to the IDFT unit 260, so that the amount of calculation is significantly reduced. As a result, the signal amplitude can be analyzed with a small amount of calculation.

In this embodiment, the delay units 231-2
Although the 3N and the filters 241 to 24N are configured as individual components, the operation of the delay units 231 to 23N can be included in the filters 241 to 24N, respectively. This is a configuration based on another mathematical expression development which is a modified example of the mathematical expression development performed in the description of the present embodiment. As described above, it is needless to say that the same processing as that of the present embodiment can also be realized by various other configurations based on mathematical formula deformation. In other words, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical formula deformation.

Further, in the present embodiment, the signal analyzer is embodied in a mathematically strict form, but a configuration exceeding practical specifications is not always necessary. For example, the spectrum inverters 272 to 27N are necessary to perform mathematically strict processing, but if the purpose is merely to analyze the amplitude value, no problem occurs even if omitted.

(Second Embodiment) FIG. 5 shows a second embodiment of the present invention.
It is a block diagram showing the composition of the signal analysis device concerning an embodiment. As shown in FIG. 5, the signal analyzer includes an input terminal 110, an analysis filter bank 200, phase detectors 321 to 32N, and output terminals 401 to 40N. In FIG. 5, the same components as those in the first embodiment shown in FIG. 1 are denoted by the same reference numerals.

Hereinafter, the operation of the signal analyzer according to the second embodiment of the present invention will be described with reference to FIG. The first
The difference between this embodiment and the third embodiment is that the mean square means 311 to 311
Since only phase detectors 321 to 32N are provided instead of 1N, detailed description of the operation of the other components is omitted.

As in the first embodiment, the analysis filter bank 200 extracts a signal for each narrow band from a signal input via the input terminal 110. The extracted narrowband signal is input to each of the phase detectors 321 to 32N, and the phase detector 321 to 32N detects the phase of the signal for each narrowband. The detected phase of each narrowband signal is output from the signal analyzer through output terminals 401 to 40N.

Here, as described in the first embodiment, each narrow-band signal output from the analysis filter bank 200 has a real component and an imaginary component. By calculating, the phase of each narrow-band signal is obtained.

As described above, according to the second embodiment,
Processing equivalent to omitting unnecessary arithmetic processing can be realized, and the well-known fast Fourier transform algorithm can be applied to the IDFT unit 260. The phase of the signal can be analyzed with the amount of calculation.

Note that, similarly to the first embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations.
Similarly, a configuration that exceeds practical use is not necessarily required.

(Third Embodiment) FIG. 6 shows a third embodiment of the present invention.
It is a block diagram which shows the structure of the signal-analysis-synthesis apparatus which is the signal processing apparatus which concerns on embodiment. 6, this signal analysis / synthesis apparatus includes an input terminal 110, an analysis filter bank 200, a mapper 500, a synthesis filter bank 600, and an output terminal 700. In FIG. 6, the same components as those in the first embodiment shown in FIG. 1 are denoted by the same reference numerals.

Next, the operation of the signal analysis / synthesis apparatus according to the third embodiment of the present invention will be described with reference to FIG. In addition,
The operation from the input terminal 110 to the analysis filter bank 200 is the same as the operation described in the first embodiment, and a detailed description thereof will be omitted.

The signal to be analyzed is a signal sampled at the sampling period T and is input to the analysis filter bank 200 via the input terminal 110. The analysis filter bank 200 extracts a signal for each narrow band from the input signal. The mapper 500 performs frequency conversion for each narrowband signal by performing mapping processing on each narrowband signal. The synthesis filter bank 600 synthesizes narrowband signals obtained by mapping, and the synthesized signal is output from the signal analysis and synthesis device via the output terminal 700.

Hereinafter, the synthesis filter bank 600 will be described in more detail. FIG. 7 is a block diagram showing a configuration of a synthesis filter bank 600 according to the third embodiment of the present invention. As shown in FIG. 7, the synthesis filter bank includes input terminals 611 to 61N, spectrum inverters 272 to 27N, and a discrete inverse Fourier transformer 260.
, Multipliers 251 to 25N, and filters 241 to 24N
, Delay units 231 to 23N, a switch 620, and an output terminal 630. In FIG. 7, the analysis filter bank 20 in the first embodiment shown in FIG.
The same components as those of 0 are denoted by the same reference numerals.

In FIG. 7, multipliers 251 to 25
N and M with i as variables_i(i = 0, 1,..., N
-1), the multiplier M_iConstant E multiplied by_iIs
e^{-j (i}π^{/ 2N)}It is. Also, the filters 241 to 24N
, Each with i as a variable _i(I = 0, 1,..., N
-1), the filter H_iPass characteristic H_i(z) is the circumference
Wave number-f_B/ 2 to f_B/ 2 signal only passes through
Characteristics of the prototype filter,

(Equation 23) Where H _i (z) = H (−jz ^N ) _i . Further, the delay unit _{231~23N, D i (i = 0,1} , ..., N-1) i as variable respectively expressed, delay device D _i
Is z ⁻ⁱ .

Next, the operation of the synthesis filter bank 600 will be described. The synthesis filter bank 600 includes:
N signals from the 0th to the (N-1) th are input via the input terminals 611 to 61N. Each of the input signals is supplied to the IDFT unit 260, but the odd-numbered signals are supplied to the IDFT unit 260 after the spectra are inverted by the spectrum inverters 272 to 27N. The N signals input to the IDFT unit 260 are converted into N signals by performing a discrete inverse Fourier transform, and output respectively. The N signals output from the IDFT unit 260 are multiplied by predetermined constants in multipliers 251 to 25N, respectively, and the filters 241 to 24N
Is subjected to a predetermined filtering process, and the delay device 2
After being delayed for a predetermined time in 31 to 23N, it is provided to the N input terminals of the switch 620. The N input terminals are connected to the output terminal 6 at the switch 620 every time T.
30. Thereby, the delay units 231 to 23N
Are output from the sampling period T of 1
The two signals are output from the synthesis filter bank 600 via the output terminal 630.

FIG. 8 is a diagram showing an example of the spectrum of each narrow-band signal input to the synthesis filter bank 600 according to the third embodiment of the present invention, and FIG. 9 is a diagram showing the third embodiment of the present invention. In the embodiment, an example of a spectrum when only the spectrum of the signal of the band #n (n is an odd number) to which the odd number is assigned among the narrow band signals input to the synthesis filter bank 600 is inverted. FIG.
In order to reproduce the original signal from each narrow-band signal input to the synthesis filter bank 600, from each narrow-band signal,
What is necessary is just to extract the components of the narrow band corresponding to each signal and add them all. However, since each input narrowband signal preserves the spectrum of the original signal in the signal baseband, as shown in FIG. 8, odd-numbered bands (band # 1 and band # 3) If the narrow-band component corresponding to the signal is directly extracted from the signal of (1), the signal is extracted in a state where the spectrum is inverted with respect to the original signal.
Therefore, for the signals of the odd-numbered bands (band 1 and band 3), the spectrum is inverted in advance to change to the spectrum shown in FIG. 9, and then the components of each narrow band (shaded area)
May be extracted and added. Such a process of inverting the spectrum in advance is performed by the spectrum inverter 272.
２７27N.

Hereinafter, the synthesis filter bank 60 described above is used.
The fact that signals can be synthesized by using 0 will be described using mathematical expressions.

First, the input signal of the k-th BPF is converted to S _k (z
^N ) and the transfer function H _k (z) of the _kth BPF is

(Equation 24) Output signal Y (z)

(Equation 25) It expresses.

As described above, from the input signal S _k (z ^N ), the corresponding narrow-band signal component (the signal component indicated by oblique lines in FIG. 9) is extracted and k = 0, 1,. Since the output signal Y (z) is obtained by adding −1, the relationship between S _k (z ^N ), H _k (z) and Y (z) is as follows.

(Equation 26) here,

[Equation 27] And can be transformed,

[Equation 28] Becomes

When this is expressed in a matrix, {Y} = [H _k ]
・ {S} ... (27) However,

(Equation 29) It is.

Here, a BPF for extracting each narrowband signal
Paying attention to the fact that the digital filter H _k (z) is obtained by frequency-shifting the prototype digital filter H (z) shown in FIG. 3, the matrix is subjected to frequency conversion.

[0094] This conversion is the case of the k-th _{filter, f → f- (2f B ·} k + f B / 2) ... (8) becomes equivalent to the conversion. Here, z = e ^j ω ^T , ω = 2πf, T = 1 / (2Nf _B ) (9) Therefore, each of z and z ^N is z → z · W ^k · e ^j π ^{/ 2N} is converted ^{W = e -j2 π / N ...} (10) z N → -jz N ... (11).

Here, the transfer function H (z) of the prototype digital filter is

[Equation 30] When the frequency transformation is performed on the matrix related to the filter bank in Expression (27), [H _k ] = [H] · [E] using Expression (14), Expression (15), and Expression (16). ] · [W] (28) Similarly, output signals Y (z) and k
When frequency conversion is performed on the input signal S _k (z ^N ) to the BPF,

(Equation 31) Becomes However,

(Equation 32) It is.

From the above results, the synthesis filter bank 6
00 indicates that the signal can be synthesized. In addition,
In the synthesis filter bank 600, the mapping device 500
Are arranged in the order of band # 0, band # 1, band # 2,..., Band # (N
It is assumed that each input terminal and a circuit corresponding to each BPF are connected in the order of -1).

Next, the mapping unit 500 will be described in more detail. FIG. 10 is a diagram illustrating an example of characteristics of the mapper 500 according to the third embodiment of the present invention. FIG.
5, the horizontal axis indicates the number of the output point of the analysis filter bank 200, and the vertical axis indicates the number of the input point of the synthesis filter bank 600. The same number on the horizontal axis and the vertical axis indicates the same band.

FIG. 10 shows, as examples of the characteristics of the mapper 500, three different mappings F ₁₁ , F ₁₂ , from the N outputs of the analysis filter bank 200 to the N inputs of the synthesis filter bank 600. three characteristics which give F ₂₁ is shown. That is, in the mapping F _11, the output of the band #n obtained by the analysis filter bank 200,
It is arranged at the input of band #n of synthesis filter bank 600 (where 0 ≦ n ≦ N−1). That is, since the frequency component of the band #n output from the analysis filter bank 200 is treated as a frequency component of the band #n also in the synthesis filter bank 600, the mapping F ₁₁ is
Analysis filter bank 200 and synthesis filter bank 6
This is a mapping equivalent to directly connecting 00.

[0099] In mapping F _12, the output of the band #n analysis filter bank 200, synthesis filter bank 6
00 band # 2n (where 0 ≦ 2
n ≦ N−1). That is, the frequency components of the band #n output from the analysis filter bank 200, because it is treated as a frequency component of the band # 2n in synthesis filter bank 600, the mapping F ₁₂ is extended the frequency axis to double Mapping.

[0100] In mapping F _21, the output of band # 2n analysis filter bank 200 is placed at the input of the band #n synthesis filter bank 600 (however, 0 ≦ 2
n ≦ N−1). In other words, the frequency component of band # 2n output from analysis filter bank 200 is processed as a frequency component of band #n in synthesis filter bank 600, so that mapping F ₂₁ sets the frequency axis to １ ／.
This is a mapping that compresses twice.

The mapping device 500 makes it possible to realize a general mapping relationship including these three typical examples. However, such a mapping is the possible, the input and the output of the synthesis filter bank 600 of the analysis filterbank 200, from being represented in the baseband of all the sampling frequency 2f _B regardless of the band It is.

FIG. 11 is a block diagram showing a configuration of a mapper 500 according to the third embodiment of the present invention.
11, the mapping device 500 includes input terminals 511 to 51N, a matrix device 520, and an interpolator 5
31 to 53N and output terminals 541 to 54N.

Hereinafter, with reference to FIG.
00 will be described. In FIG. 11, narrowband signals output from the analysis filter bank 200 are input to the matrix unit 520 via input terminals 511 to 51N. The matrix unit 520 extracts two signals for each narrow band from the input signal. Interpolator 531
To 53N perform linear interpolation based on the extracted two signals to generate output signals. The generated output signal is output from the mapper 500 via the output terminals 541 to 54N.

Next, the operations of the matrix unit 520 and the interpolators 531 to 53N will be described in more detail. First, in this mapping device 500, the input point x
(X = 0, 1,..., N−1) and an output point y (y = 0, 1,.
, N-1) is represented by a mapping function x = g (y). However, x = g (y) is defined as a set including all integers y from 0 to N−1, and 0 ≦ x <N−
This is a function whose range is a set including only a real number x of 1.
Here, the input point x indicates the input position of the signal to the mapper 500, and corresponds to the output point of the analysis filter bank 200 on a one-to-one basis. The signal of the band #x is input from the input point x. The output point y here indicates the output position of the signal from the mapper 500, and corresponds one-to-one with the input point of the synthesis filter bank 600, and is output from the output point y. The signal is input to synthesis filter bank 600 as a signal of band #y.

[0105] Here, an operation when obtaining a second output MO _a. First, the matrix device 520
Based on the above mapping function, the input MI _i (i = 0,1,
.., N−1), two consecutively-numbered inputs MI _ia and MI _{ia + 1} are extracted. Here, ia is ia = [g
(A)] (where [m] represents the largest integer not exceeding m). Interpolator 531~53N outputs MI _ia and MI _{ia + 1} in the linear interpolation performed based output MO _a. This interpolation operation is represented by the following equation. MO _a = MI _ia + α (MI _{ia + 1} −MI _ia ) α = g (a) − [g (a)]

[0106] To illustrate the above processing in more detail, for example, in FIG. 10, when the mapper 500 has a characteristic mapping F _12, is input to the input point 3 of the synthesis filter bank 600 values think of. In this case, the output to be obtained is MO ₃ , and the two inputs MI _ia and MI _{ia + 1} extracted by the matrix unit 520 are MI ₁ and MI ₂ , respectively, from the characteristics of the mapping F ₁₂ . Then, these interpolators 531 to 53N output these 2
The linear interpolation of two inputs is performed. In this case, since α = g (a) − [g (a)] = 1.5−1 = 0.5, MO ₃ = MI ₁ +0.5 (MI ₂ −MI ₁ ), and the desired output is obtained.

As described above, in the third embodiment, the analysis filter bank 200 and the synthesis filter bank 6
00, it is possible to realize processing equivalent to omitting unnecessary calculation processing, and to apply a well-known fast Fourier transform algorithm to the IDFT unit 260, so that the amount of calculation is significantly reduced. Thus, signal analysis and synthesis can be performed with a small amount of calculation. In the mapper 500, by setting the connection relationship between the output from the analysis filter bank 200 and the input to the synthesis filter bank 600 by mapping,
Each of the frequency components of the input signal can be easily converted to different frequency components. In these operations, since the input signal is processed without being compressed or expanded in the time axis direction, it is possible to perform signal frequency conversion without changing the time length.

Note that, for the same reason as in the first embodiment, the present embodiment is not limited to the above-described configuration, but may take various other configurations based on mathematical mathematical transformations. Further, an example of the characteristic of the mapping device 500 is 1
Although expressed by the following function, it is not limited to the linear function. Furthermore, although the interpolation in the interpolators 531 to 53N is the primary interpolation, it goes without saying that the interpolation can be omitted or a higher-order interpolation can be selected according to the required output quality.

(Fourth Embodiment) FIG. 12 is a block diagram showing a configuration of a signal analysis / synthesis apparatus which is a signal processing apparatus according to a fourth embodiment of the present invention. As shown in FIG. 12, the signal analyzing and synthesizing device includes a first input terminal 110
, A second input terminal 120 and an analysis filter bank 2
00, a variable mapper 800, a synthesis filter bank 600, and an output terminal 700. In addition,
12, the same components as those in the third embodiment shown in FIG. 6 are denoted by the same reference numerals.

The operation of the signal analysis / synthesis apparatus according to the fourth embodiment of the present invention will be described below with reference to FIG. The only difference between the third embodiment and the third embodiment is that a variable mapper 800 and a second input terminal 120 for inputting time information are provided instead of the mapper 500. Detailed description of the operation of the configuration is omitted.

The signal to be analyzed is a signal sampled at the sampling period T, and is input to the analysis filter bank 200 via the input terminal 110. The time information is input to the variable mapper 800 via the second input terminal 120. The analysis filter bank 200
A signal is extracted for each narrow band from the input signal. The variable mapper 800 performs frequency conversion of each narrowband signal for each narrowband by variable mapping processing based on the input time information. Synthesis filter bank 600
Synthesizes the narrowband signals obtained by the mapping, and outputs the synthesized signal from the signal analysis and synthesis device via the output terminal 700.

Hereinafter, the variable mapper 800 will be described in more detail. FIG. 13 is a block diagram illustrating a configuration of a variable mapper 800 according to the fourth embodiment of the present invention. As shown in FIG. 13, the variable mapping device 800 includes a first input terminal 511 to 51N, a second input terminal 810, a variable matrix device 820, interpolators 531 to 53N, and output terminals 541 to 54N. And In FIG. 13, the same components as those in the mapper 500 shown in FIG. 11 are denoted by the same reference numerals.

Next, the operation of the variable mapper 800 will be described with reference to FIG. In FIG. 13, each narrow-band signal output from the analysis filter bank 200 is
The signals are input to the variable matrix device 820 via the first input terminals 511 to 51N. Also, the second input terminal 120
Is also input through the second input terminal 810.
Is input to the variable matrix unit 820 via The variable matrix device 820 includes first input terminals 511 to 51
From the signal input via N, two signals are extracted for each narrow band based on time information. Interpolators 531 to 53
Each of N performs linear interpolation based on the two extracted signals to generate an output signal. The generated output signal is output from variable mapper 800 via output terminals 541 to 54N.

Next, the above-described operations of the variable matrix unit 820 and the interpolators 531 to 53N will be described in more detail. First, in the variable mapping device 800,
The input point x (x = 0, 1,..., N−1) and the output point y (y =
0, 1,..., N−1) is a mapping function x = g (y,
t) (where t is a value indicating the time information input via the second input terminal 120). Note that the input point x here is a variable mapping unit 8
00 indicates the input position of the signal to the output filter 00 and corresponds to the output point of the analysis filter bank 200 on a one-to-one basis. Is entered. The output point y here indicates the output position of the signal from the variable mapper 800 and corresponds one-to-one with the input point of the synthesis filter bank 600, and is output from the output point y. Is input to the synthesis filter bank 600 as a signal of the band #y.

[0115] Here, an operation when obtaining a second output MO _a. First, the variable matrix unit 820
Is based on this mapping function based on the input MI _i (i = 0,
1,..., N−1), two input numbers MI _{ia that} are consecutive in number.
And MI _{ia + 1} . Here, ia is ia =
[G (a, t)] (where [m] represents the largest integer not exceeding m). The interpolators 531 to 53N output the MI _ia
And MI _{ia + 1} in based performing linear interpolation and outputs an output MO _a. This interpolation operation is represented by the following equation. MO _a = MI _ia + α (MI _{ia + 1} −MI _ia ) α = g (a, t) − [g (a, t)]

As described above, such a mapping is possible because the output of the analysis filter bank 200 and the input of the synthesis filter bank 600 are all equal to the sampling frequency 2f _B regardless of the band. This is because it is expressed in baseband.

As described above, in the fourth embodiment, the analysis filter bank 200 and the synthesis filter bank 6
00, it is possible to realize processing equivalent to omitting unnecessary calculation processing, and to apply a well-known fast Fourier transform algorithm to the IDFT unit 260, so that the amount of calculation is significantly reduced. Thus, signal analysis and synthesis can be performed with a small amount of calculation. In the variable mapper 800, the input relationship is dynamically changed depending on time by dynamically changing the connection relationship between the output from the analysis filter bank 200 and the input to the synthesis filter bank 600 by the variable mapping. Each of the frequency components of the signal can be converted to different frequency components by a time-varying conversion. Such dynamic conversion of frequency components is very effective as an effect for electronic devices and the like. In these operations, since the input signal is processed without being compressed or expanded in the time axis direction, it is possible to perform signal frequency conversion without changing the time length.

Note that, like the first embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations.
Further, similarly to the third embodiment, the variable mapper 80
Although the interpolation in the 0 interpolators 531 to 53N is the primary interpolation, it goes without saying that the interpolation can be omitted or the higher-order interpolation can be selected according to the required output quality.

(Fifth Embodiment) FIG. 14 is a block diagram showing a configuration of a signal analysis / synthesis apparatus which is a signal processing apparatus according to a fifth embodiment of the present invention. As shown in FIG. 14, the signal analyzing / synthesizing device includes a first input terminal 110
, Third input terminals 131 to 13N, an analysis filter bank 200, a variable amplitude adjuster 900, a synthesis filter bank 600, and an output terminal 700.
In FIG. 14, the same components as those in the third embodiment shown in FIG. 6 are denoted by the same reference numerals.

The operation of the signal analysis / synthesis apparatus according to the fifth embodiment of the present invention will be described below with reference to FIG. The configuration of the third embodiment is different from that of the third embodiment only in that a variable amplitude adjuster 900 and third input terminals 131 to 13N for inputting amplitude adjustment values are provided instead of the mapper 500. Detailed description of the operation of the other components is omitted.

The signal to be analyzed is a signal sampled at the sampling period T, and is input to the analysis filter bank 200 via the input terminal 110. The amplitude adjustment value is input to the variable amplitude adjuster 900 via the third input terminals 131 to 13N. Analysis filter bank 2
00 extracts a signal for each narrow band from the input signal. The variable amplitude adjuster 900 adjusts the amplitude of each narrowband signal based on the input amplitude adjustment value. The synthesis filter bank 600 synthesizes the narrow-band signals whose amplitudes have been adjusted, and the synthesized signal is output from the signal analysis and synthesis device via the output terminal 700.

Hereinafter, the variable amplitude adjuster 900 will be described in more detail. FIG. 15 is a block diagram showing a configuration of the variable amplitude adjuster 900 according to the fifth embodiment of the present invention. As shown in FIG. 15, this variable amplitude adjuster 900
Has first input terminals 511 to 51N, third input terminals 911 to 91N, amplitude calculators 921 to 92N, and output terminals 541 to 54N. In FIG. 15, the same components as those of the mapping device 500 shown in FIG. 11 are denoted by the same reference numerals.

Next, the operation of the variable amplitude adjuster 900 will be described with reference to FIG. As shown in FIG. 15, the narrowband signals output from the analysis filter bank 200 are input to amplitude calculators 921 to 92N via first input terminals 511 to 51N, respectively. Similarly, the amplitude adjustment values input via the third input terminals 131 to 13N are also input to the amplitude calculators 921 to 92N via the third input terminals 911 to 91N, respectively. Amplitude calculator 92
Each of 1 to 92N changes the amplitude of the signal input via the first input terminal 511 to 51N based on the amplitude adjustment value. The signals whose amplitudes have been changed are output to the variable amplitude adjusters 9 through output terminals 541 to 54N.
Output from 00.

As described above, the present embodiment operates as a graphic equalizer that can arbitrarily change the amplitude of an input signal for each narrow band by setting an amplitude adjustment value.

As described above, in the fifth embodiment, the analysis filter bank 200 and the synthesis filter bank 6
00, it is possible to realize processing equivalent to omitting unnecessary calculation processing, and to apply a well-known fast Fourier transform algorithm to the IDFT unit 260, so that the amount of calculation is significantly reduced. Thus, signal analysis and synthesis can be performed with a small amount of calculation.

Further, in the present embodiment, the third and fourth
Similar to the embodiment, the input and the output of the synthesis filter bank 600 of the analysis filterbank 200 is represented in the baseband of all the sampling frequency 2f _B regardless of the band. Therefore, since the sampling frequency at the time of processing by the amplitude calculators 921 to 92N is reduced to 1 / N with respect to the sampling frequency of the signal to be analyzed, the graphic equalizer processing is realized with a very small amount of calculation. can do.

Note that, like the first embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations.

Next, FIGS. 1, 16 (corresponding to FIG. 2), FIG.
7, (corresponding to FIG. 3), FIG. 18 (corresponding to FIG. 4), FIGS. 5, 6, 19 (corresponding to FIG. 7), FIG. 20 (corresponding to FIG. 8), FIG. Sixth to ninth embodiments will be described with reference to FIGS. 10, 14 and 15. In the sixth to ninth embodiments, the characteristics of the digital filter and the constant multiplied by the signal passed through the digital filter are different from those in the first to fifth embodiments (the sixth embodiment). The form is the first embodiment, the seventh embodiment is the second embodiment, the eighth embodiment is the third embodiment, and the ninth embodiment is the fifth embodiment. Respectively.)

(Sixth Embodiment) The configuration of a signal analyzer according to a sixth embodiment of the present invention is shown in FIG.
As shown in FIG. 1, this signal analyzer has an input terminal 11
0, an analysis filter bank 200, and a root mean square 31
1 to 31N and output terminals 401 to 40N.

With reference to FIG. 1, the operation of the signal analyzer according to the sixth embodiment of the present invention will be described. The signal to be analyzed is input via an input terminal 110 to the filter bank 20 for analysis.
Input to 0. The analysis filter bank 200 is a filter circuit for band division, divides a frequency band of an input signal into N narrow bands, and extracts a signal for each narrow band from the input signal. Mean squares 311 to 31N
Respectively detect the amplitude of the signal extracted for each narrow band. The amplitude of each detected narrowband signal is
The signals are output from the signal analyzer through output terminals 401 to 40N, respectively.

Hereinafter, the analysis filter bank 200 will be described in more detail. In the following, it is assumed that a signal input to the analysis filter bank 200 as a signal to be analyzed is a signal sampled at a sampling period T. Further, it is assumed that all N narrow-bandwidths f _B obtained by dividing the frequency band of the input signal to be analyzed are equal. The sampling frequency f _s = 1 / T of the bandwidth f _B and the input signal, in the present embodiment, it is assumed to satisfy the following relationship. Nf _B = f _s These assumptions apply to other embodiments described later.

FIG. 16 is a block diagram showing a configuration of an analysis filter bank 200 according to the sixth embodiment of the present invention. As shown in FIG. 16, the analysis filter bank 200 includes an input terminal 210, a switch 220, delay units 231 to 23N, filters 241 to 24N, multipliers 251 'to 25N', and a discrete inverse. Fourier transformer 26
0, the spectrum inverters 272 to 27N, and the output terminal 2
81-28N.

When the delay units 231 to 23N are represented by D _i (i = 0, 1,..., N−1) using i as a variable, the delay amount in the delay unit D ₀ is z ⁻⁰ , Delay device D _i
The delay amount at (i ≠ 0) is z ^{− (Ni)} . Further, the filter 241～24N, i as variable respectively H _i
(I = 0, 1,..., N−1), the pass characteristic of the filter _Hi , that is, the transfer function H _i (z) is expressed by the frequency −f _B /
The characteristic, ie, transfer function, of a filter (hereinafter, referred to as a “prototype filter”) through which only signals from 2 to f _B / 2 pass

[Equation 33] Where H _i (z) = H (−z ^N ) _i . _{Further, M i (i = 0,1,} ..., N-1) multiplier 251'～25N ', i as variable respectively expressed with constant E _i is multiplied in multiplier M _i is e ^{-j (} a ^{i π / N).}

Next, the operation of the analysis filter bank 200 according to the sixth embodiment of the present invention will be described with reference to FIG.

The signal input to the analysis filter bank 200 is supplied to the switch 220 via the input terminal 210. The switch 220 sets the output destination of the supplied signal to the delay units 231, 232,..., 23N, 231, 232,.
, In a cyclic manner every time T. That is, a signal sampled at the sampling period NT is input to each of the delay units 231 to 23N. Delay device 2
The signals input to 31 to 23N are respectively supplied to filters 241 to 24N after being delayed by a predetermined time. Here, the delay amount z ^-1 is a time corresponding to the sampling period T of the original signal. The signals supplied to the filters 241 to 24N are respectively subjected to a predetermined filtering process, and thereafter, are multiplied by predetermined constants in multipliers 251 ′ to 25N ′. The signals multiplied by the constants in the multipliers 251 'to 25N' are subjected to discrete inverse Fourier transform by an N-point discrete inverse Fourier transformer (hereinafter referred to as "IDFT unit") 260. Odd-numbered outputs of the 0-th to (N-1) -th outputs of the IDFT unit 260 are added to the spectrum inverters 272 to 27N every -1 sample.
Is multiplied. The even-numbered outputs of the IDFT unit 260 and the outputs of the spectrum inverters 272 to 27N are output from the analysis filter bank 200 via output terminals 281 to 28N corresponding to the respective outputs.

Hereinafter, the principle of frequency analysis in the signal analyzer according to the present embodiment, that is, the fact that the frequency analysis of a signal can be performed by the operation of the above-described signal analyzer will be described with reference to the drawings and mathematical formulas.

In general, when a signal sampled at the sampling frequency f _s is equally divided by N BPFs for every 1 / N of the entire frequency band of the signal and output, the output signal is Have only 1 / N the bandwidth of the original signal. Therefore, N B
When focusing on each of the PFs, the sampling frequency can be reduced to f _s / N by the sampling theorem without impairing the information amount of the output signal of each BPF. This operation of lowering the sampling frequency is thinning. The principle of this thinning will be described with reference to FIGS. In the following, the N narrow bands obtained by equally dividing the entire frequency band of the input signal to the analysis filter bank 200 are divided into bands # 0, # 1, # 2,. , Band # (N−1) (the same applies to other embodiments). As shown in FIGS. 17 and 18, the 0th BP
The band extracted by F is extracted from band # 0, the band extracted by the first BPF is extracted by band # 2, the band extracted by the second BPF is extracted by band # 3, and the band extracted by the third BPF. The bands correspond to band # 1, respectively.

FIG. 17 is a diagram showing an example of the frequency spectrum of each narrow-band signal output from the BPF. FIG.
FIG. 8 is a diagram showing an example of a frequency spectrum after thinning out each narrow band signal output from the BPF. The signal sampled at the sampling frequency f _s has four B
PF (hereinafter referred to as "0th to 3rd BPF")
When the signal is equally divided and output for each quarter band of the entire frequency band (that is, when the entire frequency band is divided into four narrow bands), each of the output narrow bands is output. The frequency spectrum of the signal is the spectrum shown in FIG. Further, the frequency spectrum of each narrow-band signal when the sampling frequency is reduced to f _s / 4, that is, when thinning is performed on each of the narrow-band signals output in this way, is the spectrum shown in FIG. .

[0139] Comparing Figure 17 and Figure 18, the spectrum of the signal output from the BPF in the case of FIG. 17 are stored in the signal base band (frequency 0 to F _B) in the case of FIG. 18 You can see that it is. However,
Signal of band # 1 output from third BPF and 2
As for the signal of band # 3 output from the second BPF, the spectrum in the signal base band after thinning out those signals is inverted, so that the spectrum before thinning is further reduced by inverting the spectrum. Stored in the baseband. As is well known, the signal spectrum can be inverted by multiplying the signal by -1 every other sample. As described above, even if the signals output from the BPF are thinned out, the spectra of the signals, that is, the properties of the signals are stored in the signal base band. The spectrum inversion processing is performed by the spectrum inverters 272 to 27
N.

In the above description, a signal having a spectrum as shown in FIG. 18 is extracted by thinning out the signal output from the BPF. Here, as described above, performing the thinning means changing the sampling frequency from f _s to f _s.
/ N, that is, taking out samples every Nth sample. Therefore, the remaining (N-1) samples are unnecessary. The amount of calculation for calculating the unnecessary samples increases as N increases, and as a result, the operation circuit becomes complicated and the cost increases. Therefore, in the present embodiment, of a plurality of operations corresponding to a series of operations of thinning out the signal output from the BPF, only the minimum necessary operation is performed by one filter bank, so that the operation circuit We are trying to simplify it.

Hereinafter, processing corresponding to a series of operations of extracting signals in each narrow band by the BPF and then thinning them out is as follows.
The use of the analysis filter bank 200 according to the sixth embodiment will be described using mathematical expressions to show that the analysis can be performed with only the minimum required operations.

First, the input signal F (z) is

(Equation 34) And the transfer function H _k (z) of the _kth BPF is

(Equation 35) And the output signal T _k (z) from the k-th BPF is

[Equation 36] It expresses.

Expression (1) generally indicates that the input signal F (z) is
This shows that the samples constituting the input signal F (z) can be expressed as the sum of N sample sequences (0 sequence to N-1 sequence) taken out every Nth sample. The input signal F (z) is input in the order of..., N−2, N−1, 0, 1, 2,. Equation (2) indicates that the transfer function H _k (z) of the K-th BPF is generally expressed by N partial transfer functions (0 th to N th order) obtained by taking out operations of a plurality of orders in the BPF at every Nth order. (-1st order partial transfer function). Generally, the band extracted by the k-th BPF corresponds to band # k + 1.
In the analysis filter bank 200, based on such a correspondence, the arrangement order of the output terminals of the signals extracted for each narrow band is band # 0, band # 1, band # 2,.
It is assumed that a circuit corresponding to each BPF and each output terminal are connected in the order of (N-1). Equation (3) generally represents the output signal T from the k-th BPF.
_k (z) can be expressed as the sum of N sample sequences (0 sequence to N-1 sequence) obtained by extracting every N samples constituting the output signal T _k (z) from the k-th BPF. Indicates that you can do it. Note that the sequence of the input signal F (z) and k
This corresponds to the sequence of the output signal T _k (z) from the BPF, and for example, the signal output when a 0-sequence signal is input is a 0-sequence.

The relationship between F (z), H _k (z) and T _k (z) is: T _k (z) = H _k (z) F (z)

(37) Becomes When this is expressed as a matrix,

(38) Becomes

In the equation (6), the output signal T _k (z)
Is a signal composed of N sample sequences ranging from 0 sequence to (N-1) sequence. According to the above-described sampling theorem, the properties of the signal can be preserved only by taking out one sample every N samples. That is, among the output signals in the equation (6),
It suffices that only T _k (z ^N ) ₀ is obtained by thinning. In this way, for each narrow band, only the output signal after being decimated is taken out, and a matrix expression is newly made for all the narrow bands.

[Equation 39] Becomes The matrix relating to H in the equation (7) represents a transfer function of a filter bank that performs processing by reducing the amount of calculation in consideration of thinning.

Here, paying attention to the fact that the digital filter H _k (z) for extracting each narrow-band signal can be obtained by shifting the frequency of the prototype digital filter H (z) shown in FIG. Perform frequency conversion.

[0147] This conversion is the case of the k-th _{filter, f → f- (f B ·} k + f B / 2) ... (31) becomes equivalent to the conversion. Here, z = e ^j ω ^T , ω = 2πf, T = 1 / (Nf _B ) (32) Therefore, each of z and z ^N is z → z · W _k · e ^j π ^{/ N , W = e -j2 π / N} ... (33) is converted z ^N → -z ^N ... (34).

Here, the transfer function H (z) of the prototype digital filter is

(Equation 40) And the matrix for the filter bank in equation (7) is:
When frequency conversion is performed,

[Equation 41] Becomes Here, the matrix [W] shown in Expression (14) represents a calculation process corresponding to one IDFT of N points.

From the above results, the arithmetic expression taking the thinning into account is:

(Equation 42) Becomes

As described above, the process of extracting a signal for each narrow band and then decimating the signal is performed by calculating the matrix [H], calculating the matrix [E], and calculating the matrix [W] (N-point IDFT).
Can be replaced with a series of processes combining.

Next, the operation amount of the matrix [H] will be considered. For example, if the number of band divisions N is 4 and the transfer function H (z) shown in equation (12) is realized by a 12th-order FIR (non-feedback type) filter, H (z) becomes H (z) ^{_{= a 0 + a 1 z -1}} + a 2 z -2 + a 3 z -3 + a 4 z -4 + a 5 z -5 + a 6 z -6 + a 7 z -7 + a 8 z -8 + a 9 z -9 + a 10 _{^{z -10 + a 11 z -11 +}} a 12 z -12 = (a 0 + a 4 z -4 + a 8 z -8 + a 12 z -12) + (a 1 z -1 + a 5 z -5 + a 9 z -9 ) + (A ₂ z ⁻² + a ₆ z ⁻⁶ + a ₁₀ z ⁻¹⁰ ) + (a ₃ z ⁻³ + a ₇ z ⁻⁷ + a ₁₁ z ⁻¹¹ ), so that H (z ⁴ ) ₀ = a ^{_{0 + a 4 z -4 + a}} 8 z -8 + a 12 z -12 H (z 4) 1 z -1 = a 1 z -1 + a 5 z -5 + a 9 z -9 H (z 4) 2 z -2 = a _{^{a 2 z -2 + a 6 z}} -6 + a 10 z -10 H (z 4) 3 z -3 = a 3 z -3 + a 7 z -7 + a 11 z -11. From this, when H (z) is an FIR filter, the amount of calculation of the matrix [H] is at most one filter H (z).
Therefore, according to the analysis filter bank 200 shown in FIG. 16, the calculation amount is greatly reduced, and the effect is particularly enhanced as the number N of band divisions increases.

As described above, the analysis filter bank 200 shown in FIG. 16 operates, and performs processing equivalent to simultaneous extraction of narrowband signals by the BPF and sampling thinning. As described above, in the analysis filter bank 200, the arrangement order of the output terminals of the signals extracted for each narrow band is band # 0, band # 1, band # 2,.
The circuits corresponding to the respective BPFs and the respective output terminals are connected in the order of band # (N-1).

Next, in the sixth embodiment shown in FIG. 1, the operation after each narrow-band signal is obtained by the analysis filter bank 200 will be described in more detail. k
The transfer function H _k (z) of the BPF is
(z) is derived by frequency conversion, so that band 0 to
It does not have the pass band to -f _s / 2. Thus, the output processed by H _k (z) has no component band 0~-f _s / 2, the so-called analysis signal (Analitic Signa
1). The analytic signal has a real number component and an imaginary number component, and the mean square of those components represents the amplitude. Therefore, by calculating the mean square of the real and imaginary components of the output signal of the analysis filter bank 200 in the root mean squares 311 to 31N, the amplitude of each narrow band signal can be obtained.

As described above, according to the sixth embodiment,
Processing equivalent to omitting unnecessary arithmetic processing can be realized, and a well-known fast Fourier transform algorithm can be applied to the IDFT unit 260, so that the amount of calculation is significantly reduced. As a result, the signal amplitude can be analyzed with a small amount of calculation.

In this embodiment, the delay units 231-2
Although the 3N and the filters 241 to 24N are configured as individual components, the operation of the delay units 231 to 23N can be included in the filters 241 to 24N, respectively. This is a configuration based on another mathematical expression development which is a modified example of the mathematical expression development performed in the description of the present embodiment. As described above, it is needless to say that the same processing as that of the present embodiment can also be realized by various other configurations based on mathematical formula deformation. In other words, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical formula deformation.

Further, in the present embodiment, the signal analyzer is embodied in a mathematically strict form, but a configuration exceeding practical specifications is not necessarily required. For example, the spectrum inverters 272 to 27N are necessary to perform mathematically strict processing, but if the purpose is merely to analyze the amplitude value, no problem occurs even if omitted.

(Seventh Embodiment) The configuration of a signal analyzer according to the seventh embodiment of the present invention is shown in FIG.
As shown in FIG. 5, this signal analyzer has an input terminal 11
0, the analysis filter bank 200, and the phase detector 32
1 to 32N and output terminals 401 to 40N. In FIG. 5, the same components as those in the sixth embodiment shown in FIG. 1 are denoted by the same reference numerals.

Hereinafter, the operation of the signal analyzer according to the seventh embodiment of the present invention will be described with reference to FIG. The sixth
The difference between this embodiment and the third embodiment is that the mean square means 311 to 311
Since only phase detectors 321 to 32N are provided instead of 1N, detailed description of the operation of the other components is omitted.

As in the sixth embodiment, the analysis filter bank 200 extracts a signal for each narrow band from a signal input through the input terminal 110. The extracted narrowband signal is input to each of the phase detectors 321 to 32N, and the phase detector 321 to 32N detects the phase of the signal for each narrowband. The detected phase of each narrowband signal is output from the signal analyzer through output terminals 401 to 40N.

Here, as described in the sixth embodiment, each narrow-band signal output from the analysis filter bank 200 has a real component and an imaginary component. By calculating, the phase of each narrow-band signal is obtained.

As described above, according to the seventh embodiment,
Processing equivalent to omitting unnecessary arithmetic processing can be realized, and the well-known fast Fourier transform algorithm can be applied to the IDFT unit 260. The phase of the signal can be analyzed with the amount of calculation.

Note that, similarly to the sixth embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations.
Similarly, a configuration that exceeds practical use is not necessarily required.

(Eighth Embodiment) The configuration of a signal analysis / synthesis apparatus which is a signal processing apparatus according to an eighth embodiment of the present invention is shown in FIG. 6, this signal analysis / synthesis apparatus includes an input terminal 110, an analysis filter bank 200, a mapper 500, a synthesis filter bank 600, and an output terminal 700. In FIG. 6, the same components as those in the sixth embodiment shown in FIG. 1 are denoted by the same reference numerals.

Next, the operation of the signal analysis / synthesis apparatus according to the eighth embodiment of the present invention will be described with reference to FIG. In addition,
The operation from the input terminal 110 to the analysis filter bank 200 is the same as the operation described in the sixth embodiment, and a detailed description will be omitted.

The signal to be analyzed is a signal sampled at the sampling period T, and is input to the analysis filter bank 200 via the input terminal 110. The analysis filter bank 200 extracts a signal for each narrow band from the input signal. The mapper 500 performs frequency conversion for each narrowband signal by performing mapping processing on each narrowband signal. The synthesis filter bank 600 synthesizes narrowband signals obtained by mapping, and the synthesized signal is output from the signal analysis and synthesis device via the output terminal 700.

Hereinafter, the synthesis filter bank 600 will be described in more detail. FIG. 19 is a block diagram showing a configuration of a synthesis filter bank 600 according to the eighth embodiment of the present invention. As shown in FIG. 19, this synthesis filter bank includes input terminals 611 to 61N, spectrum inverters 272 to 27N, and a discrete inverse Fourier transformer 2
60, multipliers 251 ′ to 25N ′, and a filter 241.
-24N, delay units 231-23N, and switch 620
And an output terminal 630. In FIG. 19, the same components as those of the analysis filter bank 200 in the sixth embodiment shown in FIG. 16 are denoted by the same reference numerals.

In FIG. 19, multipliers 251 'to 251'
25N ′ and M _i (i = 0, 1,
..., expressed N-1) and the constants E _i is multiplied in multiplier M _i is ^{e -j (i π / N)} . Also, the filters 241-2
4N and H _i (i = 0, 1,
, N-1), the pass characteristic H _i (z) of the filter H _i.
Is the characteristic or transfer function of the prototype filter, which is a filter through which only signals from frequencies −f _B / 2 to f _B / 2 pass.

[Equation 43] Where H _i (z) = H (−z ^N ) _i . Further, the delay unit _{231~23N, D i (i = 0,1} , ..., N-1) i as variable respectively expressed, delay device D _i
Is z ⁻ⁱ .

Next, the operation of the synthesis filter bank 600 will be described. The synthesis filter bank 600 includes:
N signals from the 0th to the (N-1) th are input via the input terminals 611 to 61N. Each of the input signals is supplied to the IDFT unit 260, but the odd-numbered signals are supplied to the IDFT unit 260 after the spectra are inverted by the spectrum inverters 272 to 27N. The N signals input to the IDFT unit 260 are converted into N signals by performing a discrete inverse Fourier transform, and output respectively. The N signals output from the IDFT unit 260 are multipliers 251 ′ to 25, respectively.
N ′ is multiplied by a predetermined constant, and filters 241 to 241 are multiplied.
At 24N, a predetermined filtering process is performed, and after a predetermined time delay at delay units 231 to 23N, it is provided to N input terminals of switch 620. These N input terminals are connected to the output terminal 630 at the switch 620 every time T. Thereby, the delay unit 231
The N signals output from .about.23N are output from the synthesis filter bank 600 via the output terminal 630 as one signal of the sampling period T.

FIG. 20 is a diagram showing an example of the spectrum of each narrow-band signal input to the synthesis filter bank 600 in the eighth embodiment of the present invention.
In the eighth embodiment of the present invention, only the spectrum of the signal of the band #n (n is an odd number) to which the odd number is assigned among the narrow band signals input to the synthesis filter bank 600 is inverted. FIG. 7 is a diagram illustrating an example of a spectrum in the case. In order to reproduce the original signal from each narrow-band signal input to the synthesis filter bank 600, a narrow-band component corresponding to each signal is extracted from each narrow-band signal, and all the signals are added. Good. However, since each input narrowband signal preserves the spectrum of the original signal in the signal baseband, as shown in FIG. 20, the odd-numbered bands (band # 1, band # 3) If a narrow-band component corresponding to the signal is directly extracted from the signal of (1), the signal is extracted in a state where the spectrum is inverted with respect to the original signal. Therefore, the spectrum of the signals of the odd-numbered bands (band 1, band 3) is inverted in advance,
After changing to the spectrum shown in FIG. 21, the components (hatched portions) of each narrow band may be extracted and added. Such a process of inverting the spectrum in advance is performed in the spectrum inverters 272 to 27N.

Hereinafter, the synthesis filter bank 60 described above is used.
The fact that signals can be synthesized by using 0 will be described using mathematical expressions.

First, the input signal of the k-th BPF is converted to S _k (z
^N ) and the transfer function H _k (z) of the _kth BPF is

[Equation 44] Output signal Y (z)

[Equation 45] It expresses.

As described above, from the input signal S _k (z ^N ), the corresponding narrow-band signal component (the signal component indicated by hatching in FIG. 21) is extracted and k = 0, 1,..., N -1
, The output signal Y (z) is obtained, and the relationship between S _k (z ^N ), H _k (z), and Y (z) is as follows.

[Equation 46] here,

[Equation 47] And can be transformed,

[Equation 48] Becomes

If this is expressed in a matrix, {Y} = [H _k ]
・ {S} ... (27) However,

[Equation 49] It is.

Here, a BPF for extracting each narrow-band signal
Paying attention to the fact that the digital filter H _k (z) is obtained by shifting the frequency of the prototype digital filter H (z) shown in FIG. 17, the matrix is subjected to frequency conversion.

[0175] This conversion is the case of the k-th _{filter, f → f- (f B ·} k + f B / 2) ... (31) becomes equivalent to the conversion. Here, z = e ^j ω ^T , ω = 2πf, T = 1 / (Nf _B ) (32) Therefore, each of z and z ^N is z → z · W _k · e ^j π ^{/ N , W = e -j2 π / N} ... (33) is converted z ^N → -z ^N ... (34).

Here, the transfer function H (z) of the prototype digital filter is

[Equation 50] When the frequency conversion is performed on the matrix relating to the filter bank in Expression (27), using the expressions (14), (35), and (16), [H _k ] = [H] · [E ] · [W] (28) Similarly, output signals Y (z) and k
When frequency conversion is performed on the input signal S _k (z ^N ) to the BPF,

(Equation 51) Becomes However,

(Equation 52) It is.

From the above results, it can be seen that the synthesis filter bank 6
00 indicates that the signal can be synthesized. In addition,
In the synthesis filter bank 600, the mapping device 500
Are arranged in the order of band # 0, band # 1, band # 2,..., Band # (N
It is assumed that each input terminal and a circuit corresponding to each BPF are connected in the order of -1).

Next, the mapping device 500 will be described in more detail. An example of the characteristics of the mapper 500 according to the eighth embodiment of the present invention is shown in FIG. FIG.
5, the horizontal axis indicates the number of the output point of the analysis filter bank 200, and the vertical axis indicates the number of the input point of the synthesis filter bank 600. The same number on the horizontal axis and the vertical axis indicates the same band.

FIG. 10 shows three different mappings F ₁₁ , F ₁₂ , and N from the N outputs of the analysis filter bank 200 to the N inputs of the synthesis filter bank 600 as examples of the characteristics of the mapper 500. three characteristics which give F ₂₁ is shown. That is, in the mapping F _11, the output of the band #n obtained by the analysis filter bank 200,
It is arranged at the input of band #n of synthesis filter bank 600 (where 0 ≦ n ≦ N−1). That is, since the frequency component of the band #n output from the analysis filter bank 200 is treated as a frequency component of the band #n also in the synthesis filter bank 600, the mapping F ₁₁ is
Analysis filter bank 200 and synthesis filter bank 6
This is a mapping equivalent to directly connecting 00.

[0180] In mapping F _12, the output of the band #n analysis filter bank 200, synthesis filter bank 6
00 band # 2n (where 0 ≦ 2
n ≦ N−1). That is, the frequency components of the band #n output from the analysis filter bank 200, because it is treated as a frequency component of the band # 2n in synthesis filter bank 600, the mapping F ₁₂ is extended the frequency axis to double Mapping.

[0181] In mapping F _21, the output of band # 2n analysis filter bank 200 is placed at the input of the band #n synthesis filter bank 600 (however, 0 ≦ 2
n ≦ N−1). In other words, the frequency component of band # 2n output from analysis filter bank 200 is processed as a frequency component of band #n in synthesis filter bank 600, so that mapping F ₂₁ sets the frequency axis to １ ／.
This is a mapping that compresses twice.

The mapping unit 500 makes it possible to realize a general mapping relationship including these three typical examples. However, such a mapping is the possible, the input and the output of the synthesis filter bank 600 of the analysis filterbank 200, from being represented in the baseband of all the sampling frequency 2f _B regardless of the band It is.

The configuration of the mapper 500 according to the eighth embodiment of the present invention is shown in FIG. In FIG. 11, this mapping device 500 has input terminals 511 to
51N, a matrix unit 520, and interpolators 531 to 5
3N and output terminals 541 to 54N.

Hereinafter, with reference to FIG.
00 will be described. In FIG. 11, narrowband signals output from the analysis filter bank 200 are input to the matrix unit 520 via input terminals 511 to 51N. The matrix unit 520 extracts two signals for each narrow band from the input signal. Interpolator 531
To 53N perform linear interpolation based on the extracted two signals to generate output signals. The generated output signal is output from the mapper 500 via the output terminals 541 to 54N.

Next, the above-described operations of the matrix unit 520 and the interpolators 531 to 53N will be described in more detail. First, in this mapping device 500, the input point x
(X = 0, 1,..., N−1) and an output point y (y = 0, 1,.
, N-1) is represented by a mapping function x = g (y). However, x = g (y) is defined as a set including all integers y from 0 to N−1, and 0 ≦ x <N−
This is a function whose range is a set including only a real number x of 1.
Here, the input point x indicates the input position of the signal to the mapper 500, and corresponds to the output point of the analysis filter bank 200 on a one-to-one basis. The signal of the band #x is input from the input point x. The output point y here indicates the output position of the signal from the mapper 500, and corresponds one-to-one with the input point of the synthesis filter bank 600, and is output from the output point y. The signal is input to synthesis filter bank 600 as a signal of band #y.

[0186] Here, an operation when obtaining a second output MO _a. First, the matrix device 520
Based on the above mapping function, the input MI _i (i = 0,1,
.., N−1), two consecutively-numbered inputs MI _ia and MI _{ia + 1} are extracted. Here, ia is ia = [g
(A)] (where [m] represents the largest integer not exceeding m). Interpolator 531~53N outputs MI _ia and MI _{ia + 1} in the linear interpolation performed based output MO _a. This interpolation operation is represented by the following equation. MO _a = MI _ia + α (MI _{ia + 1} −MI _ia ) α = g (a) − [g (a)]

[0187] To illustrate the above processing in more detail, for example, in FIG. 10, when the mapper 500 has a characteristic mapping F _12, is input to the input point 3 of the synthesis filter bank 600 values think of. In this case, the output to be obtained is MO ₃ , and the two inputs MI _ia and MI _{ia + 1} extracted by the matrix unit 520 are MI ₁ and MI ₂ , respectively, from the characteristics of the mapping F ₁₂ . Then, these interpolators 531 to 53N output these 2
The linear interpolation of two inputs is performed. In this case, since α = g (a) − [g (a)] = 1.5−1 = 0.5, MO ₃ = MI ₁ +0.5 (MI ₂ −MI ₁ ), and the desired output is obtained.

As described above, in the eighth embodiment, the analysis filter bank 200 and the synthesis filter bank 6
00, it is possible to realize processing equivalent to omitting unnecessary calculation processing, and furthermore, a well-known fast Fourier transform algorithm can be applied to the IDFT unit 260, so that the calculation amount is greatly reduced. Thus, signal analysis and synthesis can be performed with a small amount of calculation. In the mapper 500, by setting the connection relationship between the output from the analysis filter bank 200 and the input to the synthesis filter bank 600 by mapping,
Each of the frequency components of the input signal can be easily converted to different frequency components. In these operations, since the input signal is processed without being compressed or expanded in the time axis direction, it is possible to perform signal frequency conversion without changing the time length.

Note that, for the same reason as in the sixth embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations. Further, an example of the characteristic of the mapping device 500 is 1
Although expressed by the following function, it is not limited to the linear function. Furthermore, although the interpolation in the interpolators 531 to 53N is the primary interpolation, it goes without saying that the interpolation can be omitted or a higher-order interpolation can be selected according to the required output quality. Also, it is needless to say that each of the frequency components of the input signal can be converted into different frequency components by dynamically changing the characteristics of the mapper depending on time. Then, when converting to a different frequency component, the frequency component to be converted may be temporally changed.

(Ninth Embodiment) The configuration of a signal analysis / synthesis apparatus which is a signal processing apparatus according to a ninth embodiment of the present invention is shown in FIG. As shown in FIG. 14, this signal analysis / synthesis apparatus includes a sixth input terminal 110, third input terminals 131 to 13N, and an analysis filter bank 20.
0, a variable amplitude adjuster 900, a synthesis filter bank 600, and an output terminal 700. FIG.
In FIG. 4, the same components as those in the eighth embodiment shown in FIG. 6 are denoted by the same reference numerals.

Hereinafter, the operation of the signal analysis / synthesis apparatus according to the ninth embodiment of the present invention will be described with reference to FIG. The only difference from the eighth embodiment is that a variable amplitude adjuster 900 and third input terminals 131 to 13N for inputting amplitude adjustment values are provided instead of the mapper 500. Detailed description of the operation of the other components is omitted.

The signal to be analyzed is a signal sampled at the sampling period T, and is input to the analysis filter bank 200 via the input terminal 110. The amplitude adjustment value is input to the variable amplitude adjuster 900 via the third input terminals 131 to 13N. Analysis filter bank 2
00 extracts a signal for each narrow band from the input signal. The variable amplitude adjuster 900 adjusts the amplitude of each narrowband signal based on the input amplitude adjustment value. The synthesis filter bank 600 synthesizes the narrow-band signals whose amplitudes have been adjusted, and the synthesized signal is output from the signal analysis and synthesis device via the output terminal 700.

Hereinafter, the variable amplitude adjuster 900 will be described in more detail. The configuration of the variable amplitude adjuster 900 according to the ninth embodiment of the present invention is shown in FIG. As shown in FIG. 15, the variable amplitude adjuster 900 includes a first
Input terminals 511-51N and a third input terminal 911-
91N, amplitude calculators 921 to 92N, and output terminal 54
1 to 54N. In FIG. 15, the same components as those of the mapping device 500 shown in FIG. 11 are denoted by the same reference numerals.

Next, the operation of the variable amplitude adjuster 900 will be described with reference to FIG. As shown in FIG. 15, the narrowband signals output from the analysis filter bank 200 are input to amplitude calculators 921 to 92N via first input terminals 511 to 51N, respectively. Similarly, the amplitude adjustment values input via the third input terminals 131 to 13N are also input to the amplitude calculators 921 to 92N via the third input terminals 911 to 91N, respectively. Amplitude calculator 92
Each of 1 to 92N changes the amplitude of the signal input via the first input terminal 511 to 51N based on the amplitude adjustment value. The signals whose amplitudes have been changed are output to the variable amplitude adjusters 9 through output terminals 541 to 54N.
Output from 00.

In this way, the present embodiment operates as a graphic equalizer that can arbitrarily change the amplitude of an input signal for each narrow band by setting an amplitude adjustment value.

As described above, in the ninth embodiment, the analysis filter bank 200 and the synthesis filter bank 6
00, it is possible to realize processing equivalent to omitting unnecessary calculation processing, and furthermore, a well-known fast Fourier transform algorithm can be applied to the IDFT unit 260, so that the calculation amount is greatly reduced. Thus, signal analysis and synthesis can be performed with a small amount of calculation.

[0197] Further, also in this embodiment, like the eighth embodiment, the output and input of the synthesis filter bank 600 of the analysis filterbank 200, the baseband of all regardless of the band sampling frequency 2f _B Is represented by Therefore, since the sampling frequency at the time of processing by the amplitude calculators 921 to 92N is reduced to 1 / N with respect to the sampling frequency of the signal to be analyzed, the graphic equalizer processing is realized with a very small amount of calculation. can do.

Note that, like the sixth embodiment, the present embodiment is not limited to the above-described configuration, and may adopt various other configurations based on mathematical mathematical transformations.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a signal analyzer according to a first embodiment of the present invention (also used in a sixth embodiment).

FIG. 2 is a block diagram showing a configuration of an analysis filter bank according to the first embodiment of the present invention.

FIG. 3 is a diagram illustrating an example of a frequency spectrum of each narrow-band signal output from a BPF.

FIG. 4 is a diagram showing an example of a frequency spectrum after thinning out each narrow-band signal output from the BPF.

FIG. 5 is a block diagram showing a configuration of a signal analyzer according to a second embodiment of the present invention.

FIG. 6 is a block diagram showing a configuration of a signal analysis / synthesis device according to a third embodiment of the present invention (also used in the eighth embodiment).

FIG. 7 is a block diagram illustrating a configuration of a synthesis filter bank according to a third embodiment of the present invention.

FIG. 8 is a diagram illustrating an example of a frequency spectrum of each narrowband signal input to a synthesis filter bank according to the third embodiment of the present invention.

FIG. 9 is an example of a frequency spectrum when only the spectrum of a signal of an odd-numbered band among the narrow-band signals input to the synthesis filter bank is inverted in the third embodiment of the present invention. FIG.

FIG. 10 is a diagram illustrating an example of a characteristic of a mapper according to the third embodiment of the present invention (also used in the eighth embodiment).

FIG. 11 is a block diagram showing a configuration of a mapping device according to a third embodiment of the present invention (also used in the eighth embodiment).

FIG. 12 is a block diagram showing a configuration of a signal analysis and synthesis device according to a fourth embodiment of the present invention.

FIG. 13 is a block diagram illustrating a configuration of a variable mapper according to a fourth embodiment of the present invention.

FIG. 14 is a block diagram showing a configuration of a signal analysis / synthesis device according to a fifth embodiment of the present invention (also used in the ninth embodiment).

FIG. 15 is a block diagram showing a configuration of a variable amplitude adjuster according to a fifth embodiment of the present invention (also used in the ninth embodiment).

FIG. 16 is a block diagram showing a configuration of an analysis filter bank according to a sixth embodiment of the present invention.

FIG. 17 is a diagram showing another example of the frequency spectrum of each narrow-band signal output from the BPF.

FIG. 18 is a diagram showing another example of the frequency spectrum after thinning out each narrow band signal output from the BPF.

FIG. 19 is a block diagram showing a configuration of a synthesis filter bank according to an eighth embodiment of the present invention.

FIG. 20 is a diagram illustrating an example of a frequency spectrum of each narrow-band signal input to the synthesis filter bank in the eighth embodiment of the present invention.

FIG. 21 shows an example of a frequency spectrum when only the spectrum of a signal of an odd-numbered band among the narrow-band signals input to the synthesis filter bank is inverted in the eighth embodiment of the present invention. FIG.

FIG. 22 is a block diagram showing a configuration of a conventional signal analyzer.

[Explanation of symbols]

 110 input terminal 120 second input terminal 131-13N third input terminal 200 analysis filter bank 210 input terminal 220 switch 231-23N delay unit 241-24N filter 251-25N, 251 '~ 25N' ... Multiplier 260 ... Discrete Inverse Fourier Transformer 272-27N ... Spectrum Inverter 281-28N ... Output Terminals 31-31N ... Square Meaner 321-32N ... Phase Detector 401-40N ... Output Terminal 500 ... Mapper 511-51N ... Input terminal 520 ... Matrix device 531-53N ... Interpolator 541-54N ... Output terminal 600 ... Synthesis filter bank 611-61N ... Input terminal 620 ... Switcher 630 ... Output terminal 700 ... Output terminal 800 ... variable mapping device 810 ... second input terminal 820 Variable matrix 900 ... variable amplitude adjuster 911～91N ... third input terminal 921～92N ... amplitude calculator

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H03M 7/30 G10L 7/04 E 7/06

Claims (22)

[Claims] 1. A frequency band of an input signal sampled at a predetermined sampling period T is divided into a predetermined number N of narrow bands, and N narrow band signals corresponding to respective narrow band components in the input signal. And a N-th to N-th delay units for respectively delaying the supplied signals by a predetermined delay amount, and receiving the input signal. And one of the 0-th to N-1th delay units is selected, and the selected delay unit is connected between the 0-th and N-1th delay units. And a switch for cyclically switching for each sample value of the input signal, and from the 0th to the N-1th through which the delayed signals output from the 0th to the N-1th delays respectively pass of N digital filters and N-th to N-th N multiplications for multiplying signals after passing through the 0th to N-1th digital filters by predetermined constants, respectively Inverse Fourier performing N points of discrete inverse Fourier transforms on the multiplied 0th to N-1th signals output from the 0th to N-1th multipliers, respectively A zero to N obtained by the discrete inverse Fourier transform.
A band division filter circuit for outputting N signals up to the first number as N narrow band signals. 2. An odd-numbered signal among the 0-th to (N-1) -th N signals after the discrete inverse Fourier transform output from the discrete inverse Fourier transformer, is sampled one sample at a time. Further comprising a spectrum inverter for multiplying by "-1", and an even-numbered signal output from the discrete inverse Fourier transformer and an odd-numbered signal after being multiplied by "-1" every other sample by the spectrum inverter. N consisting of
2. The filter circuit for band division according to claim 1, wherein N signals are output as N narrow band signals. 3. The bandwidths of the N narrow bands are all equal, and the delay amount of the i-th delay unit among the N delay units is calculated by a unit delay operation corresponding to the sampling period T. If it is expressed using the child z ⁻¹ , i =
When it is ⁰ , it is z- ⁰ , and when 1 ≦ i ≦ N−1, it is z− ^{(Ni). The} bandwidth of each narrow band is f _B, and the frequency of −f _B / 2 is f _B / The characteristics of a band-pass type digital filter having a pass band up to the frequency of 2 are given by And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−jz ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. 3. The band dividing filter circuit according to claim 1, wherein the signal that has passed through the digital filter is multiplied by e ^{-j [i} π ^{/ ( 2N)]} as the constant. 4. The bandwidths of the N narrow bands are all equal, and the delay amount of the i-th delay unit among the N delay units is calculated by a unit delay calculation corresponding to the sampling period T. If it is expressed using the child z ⁻¹ , i =
When it is ⁰ , it is z- ⁰ , and when 1 ≦ i ≦ N−1, it is z− ^{(Ni). The} bandwidth of each narrow band is f _B, and the frequency of −f _B / 2 is f _B / The characteristics of a band-pass digital filter having a pass band up to the frequency 2 are given by And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. characterized by multiplying e ^{-j [i} π ^{/ N]} as the constant to the signal that has passed through the digital filter, band dividing filter circuit according to claim 1 or 2. 5. A signal analyzer for performing frequency analysis on an input signal sampled at a predetermined sampling period T, and receiving the input signal.
The band dividing filter circuit described in the paragraph, and N pieces of numbers each outputting a mean square value of a real part and an imaginary part of each value of the N narrow band signals output from the band dividing filter circuit And a root mean square. 6. A signal analyzer for performing frequency analysis on an input signal sampled at a predetermined sampling period T, and receiving the input signal.
And the inverse tangent of the ratio I / R of the real part R and the imaginary part I of each of the N narrowband signals output from the band dividing filter circuit. A signal analyzer comprising: N phase detectors each outputting a value indicated by the signal analyzer. 7. A signal analyzing apparatus for performing frequency analysis on an input signal sampled at a predetermined sampling period T, wherein the filter circuit for band division according to claim 4, which receives the input signal, N number of mean squarers each outputting a mean square value of a real part and an imaginary part of each value of the N narrow band signals output from the dividing filter circuit. Signal analyzer. 8. A signal analysis device for performing frequency analysis on an input signal sampled at a predetermined sampling period T, wherein the filter circuit for band division according to claim 4, wherein the input signal is received. N phase detectors each outputting a value indicating an arctangent of a ratio I / R of a real part R and an imaginary part I of each of the N narrowband signals output from the dividing filter circuit. And a signal analyzer. 9. A signal processing apparatus for generating a modified signal of an input signal by giving a specific change to an input signal sampled at a predetermined sampling period T, wherein the frequency band of the input signal is set to a predetermined number N And N equal to the N-1th narrowband signal corresponding to each narrowband component of the input signal, and N times the sampling period T of the input signal. A band dividing filter circuit for generating N narrow-band signals having the sampling period of N from the input signal; N input points from 0th to N-1th and Nth from 0th to N-1th Number of output points, and from the 0th to N
The -1st narrowband signal is received at each of the 0th to N-1th input points, and is received from each output point at an input point corresponding to each output point by a predetermined mapping. A mapper that outputs the narrowband signal, and signals output from the 0th to N−1th output points in the mapper correspond to the 0th to N−1th narrowband signals, respectively. A signal processing apparatus comprising: a synthesis filter circuit that generates a modified signal of the input signal by synthesizing the input signal. 10. The mapping unit may express the mapping by a function g (y) having a predetermined set including an integer y indicating the output point as a domain and a predetermined set of real numbers as a range. , A plurality of signals to be output from the j-th output point at a plurality of input points corresponding to integers near a real number g (j) among the 0-th to (N−1) -th input points. The signal processing apparatus according to claim 9, wherein the signal processing apparatus generates the signal by interpolation based on the narrowband signal and the real number g (j). 11. The apparatus according to claim 1, wherein the mapper generates a signal MO _j to be output from the j-th output point by linear interpolation represented by the following equation.
0: MO _j = MI _i + α · (MI _{i + 1} −MI _i ) i = [g (j)] α = {g (j) − [g (j)]} where j is an integer satisfying 0 ≦ j ≦ N−1, and x = g (y) is a function defining the mapping in the mapper, and represents all integers y satisfying 0 ≦ y ≦ N−1. Is a function whose range is a set including only real numbers x satisfying 0 ≦ x <N−1, where [g (j)] is the largest integer not exceeding g (j); _i is the _ith narrowband signal received at the ith input point in the mapper. 12. The mapping device receives time information from the outside, and the correspondence between the input point and the output point given by the mapping in the mapping device changes with the passage of time based on the time information. The signal processing device according to claim 9, wherein the signal processing is performed. 13. The function g, wherein the mapper defines a predetermined set of pairs of an integer y indicating the output point and a value t indicated by the time information as a domain and a predetermined set of real numbers as a range. Assuming that the mapping is represented by (y, t), a signal to be output from the j-th output point is
Real number g among the 0th to N-1st input points
The signal is generated by interpolation based on a plurality of the narrowband signals received at a plurality of input points corresponding to integers near (j, t) and the real number g (j, t). 1
3. The signal processing device according to 2. 14. The apparatus according to claim 1, wherein the mapper generates a signal MO _j to be output from the j-th output point by linear interpolation represented by the following equation.
3. The signal processing device according to 3: MO _j = MI _i + α · (MI _{i + 1} −MI _i ) i = [g (j, t)] α = ｛g (j, t) − [g (j, t) Where j is an integer satisfying 0 ≦ j ≦ N−1, and x = g (y, t) is a function defining the mapping in the mapping unit, and 0 ≦ y ≦ N A function that sets a set including all pairs of an integer y of −1 and a value t indicated by the time information as a domain and a set including only a real number x of 0 ≦ x <N−1 as a range, [g (j, t)] is the largest integer not exceeding g (j, t), and MI _i is the _ith narrowband signal received at the ith input point in the mapper. 15. A signal processing apparatus for generating a modified signal of an input signal by giving a specific change to an input signal sampled at a predetermined sampling period T, wherein the frequency band of the input signal is set to a predetermined number N And N equal to the N-1th narrowband signal corresponding to each narrowband component of the input signal, and N times the sampling period T of the input signal. A band dividing filter circuit for generating N narrow-band signals having the sampling period from the input signal; and receiving amplitude adjustment information from the outside; An amplitude adjustment circuit that adjusts the amplitude of the narrowband signal and outputs the 0th to N−1th narrowband signals whose amplitude has been adjusted; and an output from the amplitude adjuster. And a synthesis filter circuit for generating a modified signal of the input signal by synthesizing the 0th to N-1th narrowband signals. 16. The N-th to N-1-th amplitude adjusting circuits respectively adjust the amplitudes of the 0-th to N-1-th narrow-band signals output from the band dividing filter circuit. , Which correspond to the 0th to N−1th narrowband signals output from the band dividing filter circuit.
The Nth to N-1th adjustment values are received as the amplitude adjustment information, and if i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th amplitude adjuster is the i-th amplitude adjuster. 16. The signal processing device according to claim 15, wherein the amplitude of the i-th narrowband signal is adjusted based on the adjustment value. 17. The band splitting filter circuit includes: 0th to (N-1) th N delayers for respectively delaying supplied signals by a predetermined delay amount; The delay device receives and supplies the selected delay device to one of the 0th to N-1th delay devices, and supplies the selected delay device to the 0th to N-1th delay devices. A switch for cyclically switching between each sample value of the input signal between the 0th and N-1th N-th signals through which signals output from the 0th to N-1th delay devices pass, respectively.
N digital multipliers for multiplying signals after passing through the digital filters from 0 to N-1 by predetermined constants, respectively. And N points of discrete inverse Fourier transform are performed on the multiplied 0th to N−1th signals output from the 0th to N−1th multipliers, respectively. A discrete inverse Fourier transformer for outputting N signals from 0th to (N-1) th which are signals after the Fourier transform; and an odd number among the N signals output from the discrete inverse Fourier transformer For each signal of the １
A spectrum inverter for multiplying by 1 ”, and an odd-numbered signal obtained by multiplying an even-numbered signal output from the discrete inverse Fourier transformer by“ −1 ”every one sample by the spectrum inverter. N consisting of
The signal processing apparatus according to claim 9, wherein N signals are output as the N narrowband signals. 18. A delay amount caused by an i-th delay unit among the N delay units is represented by a unit delay operator z ⁻¹ corresponding to the sampling period T. When i = 0, z− ⁰ , and 1 ≦ i ≦ N
When -1 z ^- a ^(Ni), wherein the bandwidth of each narrowband and f _B, -f _B / 2 of the band-pass digital filter to the frequency up to f _B / 2 and pass band from a frequency The characteristic of And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−jz ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. 18. The signal processing apparatus according to claim 17, wherein the signal that has passed through the digital filter is multiplied by e- ^{j [i [} pi ^{] / ( 2N)]} as the constant. 19. A delay amount caused by an i-th delay unit among the N delay units is represented by a unit delay operator z ^-1 corresponding to the sampling period T. When i = 0, z− ⁰ , and 1 ≦ i ≦ N
When -1 z ^- a ^(Ni), wherein the bandwidth of each narrowband and f _B, -f _B / 2 of the band-pass digital filter to the frequency up to f _B / 2 and pass band from a frequency The characteristic of And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. 18. The signal processing apparatus according to claim 17, wherein the signal that has passed through the digital filter is multiplied by e- ^{j [i [} pi ^{] / N ]} as the constant. 20. The synthesis filter circuit, comprising: a spectrum inverter that multiplies a signal output from an odd-numbered output point in the mapping unit by “−1” every one sample; , And N points of discrete inverse Fourier transform are performed on N signals composed of a signal output from the output point and a signal multiplied by “−1” every other sample by the spectrum inverter. A discrete inverse Fourier transformer which outputs N signals from 0th to N-1th which are signals after the inverse inverse Fourier transform, and 0th to N-1 output from the discrete inverse Fourier transformer N-th to N-th multipliers for multiplying the up to N-th signal by a predetermined constant, respectively, N digital filters from 0th to N-1th through which the 0th to N-1th multiplied signals after output are respectively passed; and 0th to N-1th digital filters. Selecting from the 0th to N-1th N delayers for delaying the signal after passing each by a predetermined delay amount, and the 0th to N-1th delayers And outputs the delayed signal output from one of the delay units as a modified signal of the input signal, and sets the selected delay unit between the 0th to N-1th delay units. 10. A switch for cyclically switching for each sample value of the input signal.
6. The signal processing device according to 5. 21. A delay amount of an i-th delay unit among the N delay units is represented by a unit delay operator z ⁻¹ corresponding to the sampling period T, Z ^-i for an arbitrary integer i satisfying 0 ≦ i ≦ N−1, where the bandwidth of each narrow band is f _B, and the passband extends from a frequency of −f _B / 2 to a frequency of f _B / 2. The characteristic of the band-pass digital filter is And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−jz ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. ^21. The signal processing apparatus according to claim 20, wherein the signal that has passed through the digital filter is multiplied by e ^{-j [} iπ ^{/ ( 2N)]} as the constant. 22. A delay amount caused by an i-th delay unit among the N delay units is represented by a unit delay operator z ⁻¹ corresponding to the sampling period T. Z ^-i for an arbitrary integer i satisfying 0 ≦ i ≦ N−1, where the bandwidth of each narrow band is f _B, and the passband extends from a frequency of −f _B / 2 to a frequency of f _B / 2. The characteristic of the band-pass type digital filter is given by And i is any integer satisfying 0 ≦ i ≦ N−1, the i-th digital filter is
H _i (z) = H (−z ^N ) _i , where i is an arbitrary integer satisfying 0 ≦ i ≦ N−1, the i-th multiplier is the i-th multiplier. ^21. The signal processing apparatus according to claim 20, wherein the signal that has passed through the digital filter is multiplied by e- ^{j [i [} pi ^{] / N ]} as the constant.