JP2002328699A - Encoding device and decoding device - Google Patents

Abstract

(57) [Problem] To reduce the information amount of a coded sequence obtained by coding an audio signal while maintaining the sound quality of the audio signal. A coding apparatus according to the present invention calculates a mean amplitude of a frequency spectrum sequence for each of a plurality of frequency bands, thereby generating a first code representing the mean amplitude of the frequency spectrum sequence. And a coding band determining unit that quantizes a frequency spectrum sequence and determines at least one frequency band to be coded among a plurality of frequency bands, and at least one frequency band determined by the coding band determining unit. A spectrum encoding unit that generates a second code by quantizing and encoding the frequency spectrum sequence;
And a coded sequence generation unit that generates a coded sequence based on

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an encoding device and a decoding device. More specifically, the present invention relates to an encoding device and a decoding device that encode an audio signal into an encoded sequence having a small amount of information while maintaining the sound quality of the audio signal.

[0002]

2. Description of the Related Art Numerous methods have been developed for encoding and decoding audio signals including audio signals and / or music signals. In particular, recently, IS13818-7, which has been internationally standardized by ISO / IEC, has been recognized among them, and has been evaluated as a high-quality and high-efficiency encoding method. The encoding method is A
Called AC.

[0003] In recent years, AAC has been adopted in a standard called MPEG4, and a system called MPEG4-AAC having some extended functions over IS13818-7 has been defined. An example of the encoding process is INFOM
It is described in ATIVE PART.

FIG. 10 shows a configuration example of a conventional encoding apparatus 1000. The frequency spectrum sequence is encoded by the encoding device 1000
Is input to The frequency spectrum sequence is generated as described below.

[0005] The audio signal is input to a time-frequency converter (not shown) in the form of an audio discrete signal obtained by sampling the audio signal. The time-frequency converter is
The discrete signal on the time axis is converted into a spectrum on the frequency axis by orthogonal transformation or the like. In this specification, the entire spectrum on the frequency axis converted from the discrete signal on the time axis by the time-frequency conversion unit is referred to as a frequency spectrum of one frame. The frequency spectrum of one frame is divided into a plurality of frequency spectra corresponding to a plurality of frequency bands. The frequency spectrum sequence is input to the encoding device 1000.

[0006] Encoding apparatus 1000 receives a frequency spectrum sequence, and amplifies spectrum sequence by amplifying the frequency spectrum sequence by using a predetermined gain and a spectrum amplifying section 101 that generates an encoding gain obtained by encoding the predetermined gain.
0, a spectrum quantization unit 1020 that generates a quantized spectrum sequence by quantizing the amplified spectrum sequence, a Huffman coding unit 1030 that generates a Huffman-coded spectrum sequence that performs Huffman coding on the quantized spectrum sequence, An encoded sequence generation unit 1040 that generates an encoded sequence including a gain and a Huffman encoded spectrum sequence.

[0007] The spectrum amplifying unit 1010 uses a predetermined gain for a spectrum sequence corresponding to a predetermined frequency band among a plurality of frequency bands to obtain a frequency spectrum representing a frequency spectrum of the predetermined frequency band. An amplified spectrum train is generated by amplifying the spectrum train. The spectrum amplifying unit 1010 also generates an encoded gain obtained by encoding a predetermined gain.

[0008] The spectrum quantization unit 1020 generates a quantized spectrum sequence by quantizing the data of the amplified spectrum sequence by a predetermined conversion formula. In the case of the AAC method, the spectrum quantization unit 1020 quantizes the data of the amplified spectrum sequence represented by a floating-point number by rounding the data to an integer value.

The Huffman coding unit 1030 collects data of the quantized spectrum sequence several by one to generate a Huffman coded Huffman coded spectrum sequence.

[0010] The coded sequence generator 1040 transfers a coded sequence including a coded gain and a Huffman coded spectrum sequence to a decoding device (not shown).

[0011]

In recent years, it has been desired to reduce the amount of information in a coded stream obtained by coding an audio signal and to increase the compression rate of the audio signal.

[0012] In the coding apparatus 1000, the performance of information compression depends on the Huffman coding unit 1030. In the encoding device 1000, the audio signal is compressed at a high compression rate,
That is, when encoding is performed with a small amount of information, the value of the data of the quantized spectrum sequence is reduced by controlling the gain of the spectrum amplifying unit 1010, thereby reducing the amount of information encoded by the Huffman encoding unit 1030. I do.

However, if such an operation is performed, a very large number of amplitude (quantized value) values appear in the frequency spectrum obtained by decoding the Huffman-coded spectrum sequence, and a sufficient sound quality cannot be secured. There is.

The present invention has been made in view of such a conventional problem, and encodes a frequency spectrum sequence corresponding to an audio signal into a coded sequence having a small amount of information while securing the sound quality of the audio signal. It is an object of the present invention to realize an encoding device that performs decoding and an encoding device that decodes such an encoded sequence into an output spectrum sequence corresponding to a decoded audio signal.

[0015]

An encoding apparatus according to the present invention comprises:
Calculating a mean amplitude of the frequency spectrum sequence for each of the plurality of frequency bands to generate a first code representing an average amplitude of the frequency spectrum sequence; A coding band determination unit that quantizes a spectrum sequence and determines at least one frequency band to be coded; and quantizes and encodes a frequency spectrum sequence for each of the at least one frequency band determined by the coding band determination unit. By doing so, a spectrum encoding unit that generates a second code, and a coded sequence generation unit that generates a coded sequence based on the first code and the second code are provided.

[0016] The coding band determination section may determine whether the frequency spectrum sequence should be quantized and coded based on the magnitude of the first code representing the average amplitude of the frequency spectrum sequence.

[0017] The coding band determination unit is configured to generate a second one of the at least one frequency band determined to be quantized and to be coded by the spectrum coding unit.
Based on the size of the code of, from among the frequency bands that have not been determined to be quantized and encoded, the frequency band to be further quantized and encoded is re-determined, and the spectrum encoding unit performs the re-determination. By quantizing and encoding the frequency spectrum sequence for the frequency band
A second code may be generated.

A third code indicating which of the plurality of frequency bands has been quantized and encoded into a frequency spectrum sequence, the first code, and the second code; The encoded sequence generation unit may generate the encoded sequence based on the following code:

[0019] The spectrum coding section may perform Huffman coding.

[0020] The spectrum coding section may perform vector quantization.

[0021] The spectrum encoding unit may perform Huffman encoding and vector quantization.

A fourth code representing an average amplitude of the time signal sequence is obtained by calculating an average amplitude of each time signal sequence in a plurality of time domains converted into respective frequency spectrum sequences of the plurality of frequency bands. May be further provided.

At least one frequency band is divided into a plurality of sub-bands among the plurality of frequency bands which are not quantized and determined to be coded by the coding band determination unit, and are divided into a plurality of sub-bands. A sub-band gain encoding unit that generates a fifth code representing the average amplitude of each band may be further provided.

[0024] At least one of the plurality of sub-bands includes:
It may include the two or more frequency spectrum sequences.

A decoding device according to the present invention is a decoding device for decoding a coded sequence including a first code and at least one second code, wherein the first code includes a plurality of frequency codes. Each of the at least one second code is generated to represent an average amplitude of a frequency spectrum sequence for one of the bands, wherein each of the at least one second code is a frequency spectrum sequence for one of the plurality of frequency bands. And a coded sequence analysis unit for analyzing the coded sequence and detecting the first code and the at least one second code. A band gain inverse quantization unit that inversely quantizes the first code detected by the converted sequence analysis unit to an average amplitude of the frequency spectrum sequence; and a frequency band of the at least one second code, An encoding band notification section for notifying whether or not there is a corresponding frequency band to the frequency band of the first code,
In the frequency band of the at least one second code,
A spectrum inverse quantization unit that inversely quantizes and decodes the second code into the frequency spectrum sequence based on the notification of the encoding band notification unit that there is a frequency band corresponding to the frequency band of the first code. And generating a noise spectrum sequence based on the notification of the coding band notification unit that there is no frequency band corresponding to the frequency band of the first code in the frequency band of the at least one second code. A noise spectrum sequence generating unit, and an amplification unit that amplifies the frequency spectrum sequence or the noise spectrum sequence based on the average amplitude.

The coded sequence further includes a third code indicating which of the plurality of frequency bands has been quantized and coded into a frequency spectrum sequence, and the coded band notifying unit outputs a third code. Decoding a code, based on the decoded third code, whether or not there is a frequency band corresponding to the frequency band of the first code in the frequency band of the at least one second code; May be notified.

[0027] The spectrum dequantizer may perform Huffman decoding.

[0028] The spectrum inverse quantization section may perform vector inverse quantization.

The spectrum inverse quantization section may perform Huffman decoding and vector inverse quantization.

[0030] The encoded sequence further includes a fourth code representing an average amplitude of each time signal sequence in a plurality of time domains to be converted into respective frequency spectrum sequences in the plurality of frequency bands. May be further provided with a time-domain gain decoding unit that decodes the above-mentioned code into an average amplitude of the time signal sequence.

The noise spectrum sequence generating section generates a noise spectrum sequence to be converted into each of the plurality of time domain noise signals based on the fourth code decoded by the time domain gain decoding section. May be.

The encoded sequence further includes a fifth code representing an average amplitude of each of a plurality of sub-bands obtained by dividing at least some of frequency bands among the frequency bands not dequantized by the spectrum dequantizer. A sub-band decoding unit configured to decode the fifth code into an average amplitude of the sub-band, and generate a noise spectrum sequence of the plurality of sub-bands based on the decoded average amplitude. Good.

[0033]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An encoding apparatus and a decoding apparatus according to an embodiment of the present invention and a data processing system including the encoding apparatus and the decoding apparatus will be described below with reference to the drawings.

(Embodiment 1) FIG. 1 shows a configuration example of an audio signal conversion system 10 according to the present invention. The audio signal conversion system 10 includes a time-frequency conversion unit 20 that converts an audio signal into a frequency spectrum sequence, an output obtained by encoding the frequency spectrum sequence into an encoded sequence to reduce the amount of information, and decoding the encoded sequence. The system includes a data processing system 100 that generates a spectrum sequence and a frequency-time conversion unit 30 that converts an output spectrum sequence into a decoded audio signal. The decoded audio signal is reproduced by the reproducing unit 40.

The data processing system 100 includes an encoding device 110 for encoding a frequency spectrum sequence into an encoded sequence,
Decoding device 1 for decoding an encoded sequence into an output spectrum sequence
20. In the audio signal conversion system 10, the time-frequency conversion unit 20 and the encoding device 110
It functions as the transmission unit 60. Also, the decryption device 120
And the frequency-time conversion unit 30 function as the reception unit 70. The encoded sequence output from the transmitting unit 60 is temporarily recorded by an arbitrary recording unit, and is decoded and reproduced if desired. Alternatively, the coded sequence output from the transmitting unit 60 is transmitted to the receiving unit 7 via a transmission path (not shown).
Sent to 0.

The audio signal is input to the time-frequency converter 20 in the form of an audio discrete signal obtained by sampling the audio signal. The audio discrete signal is represented as a discrete signal on the time axis. Time frequency converter 20
Converts a discrete signal on the time axis into a spectrum on the frequency axis at certain time intervals. In this specification, the entire discrete signal on the time axis at a certain time interval is called a one-frame time signal, and the spectrum on the frequency axis obtained by converting the one-frame time signal is called a one-frame frequency spectrum.
The time signal of one frame is represented as a sequence of time signals of one frame. The frequency spectrum of the frame is divided into a plurality of frequency spectra corresponding to a plurality of frequency bands. In this specification, each of the plurality of divided frequency bands is called a scale factor band. A plurality of frequency spectrum data belong to each scale factor band, and each of the data is input to the encoding device 110.

The time-frequency converter 20 performs time-frequency conversion by, for example, a modified discrete cosine transform (MDCT). MDCT is a technique known to those skilled in the art. The time-frequency conversion unit 20 is provided for each specific sample (for example, 5
The time-frequency conversion is performed every 12 samples or every 1024 samples. When the number of samples of the time signal sequence is 512 and MDCT is used as the time-frequency transform, MDCT coefficients of 512 samples are obtained for each frame. In the case of using the MDCT, a description will be given later with the entire MDCT coefficient as a frequency spectrum of one frame.

FIG. 2A shows the encoding device 11 shown in FIG.
The configuration example of the encoding device 110A as an example of 0 is shown.
The encoding device 110A receives the frequency spectrum sequence and generates an encoded sequence.

The encoding device 110A calculates the average amplitude of the frequency spectrum sequence, and generates a first code representing the average amplitude of the frequency spectrum sequence.
A, a coding band determining unit 220A that quantizes a frequency spectrum sequence and determines at least one frequency band to be coded among a plurality of frequency bands, and at least one frequency band determined by the coding band determining unit 220A. Are quantized and coded for each of the frequency spectrum sequences, thereby generating a second code, a first code generated by the band gain coding unit 210A, and a spectrum coding unit 230A.
And a coded sequence generation unit 240A that generates a coded sequence based on the second code generated by A.

Band gain encoding section 210A calculates, for each scale factor band, the average amplitude of the frequency spectrum sequence belonging to the scale factor band. Calculation of the average amplitude rms of the frequency spectrum sequence for each scale factor band uses, for example, (Equation 1).

[0041]

(Equation 1) Here, sp (i) is the value of each data of the frequency spectrum sequence of the scale factor band,
n is the number of data in the frequency spectrum sequence of the scale factor band.

The band gain encoding unit 210A quantizes and encodes the average amplitude obtained for each scale factor band.

The encoded average amplitude (index) is given by, for example, (Equation 2).

[0044]

(Equation 2) Here, (int) is a function that rounds off the value below the decimal point and converts it to an integer, and log2 is a function that takes the logarithm of 2.

The quantized average amplitude (qrms) is given by, for example, (Equation 3). Here, ＾ is a function of the exponential operation.

[0046]

(Equation 3) When a frequency spectrum of one frame is divided into M frequency spectrums (composed of M scale factor bands), the quantized average amplitude is M at the maximum. The coded sequence generation unit 240A may generate the coded sequence using codes representing the average amplitudes of all M pieces. Or
The coded sequence may be generated using codes representing the average amplitude of less than M numbers from the low frequency band. Or,
A coded sequence may be generated based on a code representing one average amplitude and other information. Further, as a method of generating a coded sequence, the code calculated by (Equation 2) may be directly coded, or the difference between the average amplitudes of adjacent scale factor bands may be coded using Huffman coding or the like. Is also good.

The coding band determination section 220A determines which scale factor band belongs to the frequency spectrum string divided into a plurality of scale factor bands and quantizes and encodes it in the spectrum coding section 230A. To determine. The scale factor bands to be quantized and coded may be set in advance, for example, from low to N scale factor bands.

Here, it is assumed that a frequency spectrum sequence belonging to the N scale factor bands from the low band out of the M scale factor bands is set in advance to be encoded. Here, M and N are both natural numbers, and M is equal to or larger than N. Here, the encoding is set to be performed from the low frequency band because the influence of the low frequency spectrum is more important to human hearing when an audio signal is reproduced.

The spectrum encoding section 230A quantizes and encodes a frequency spectrum sequence belonging to the scale factor band determined to be quantized and encoded by the encoding band determining section 220A. Spectrum encoding section 230
A may use Huffman coding or may use vector quantization. Alternatively, both Huffman coding and vector quantization may be used. here,
It is assumed that the coding type of spectrum coding section 230A is specified in advance. However, the present embodiment is not limited to this, and spectrum coding section 230A outputs information representing the type of quantization and coding of the frequency spectrum sequence to coding sequence generating section 240A, and performs coding sequence generating section 2A.
40A may include that information in the coded sequence.

[0050] Coded sequence generating section 240A generates a coded sequence based on the average amplitude generated by band gain coding section 210A and the coded spectrum sequence generated by spectrum coding section 230A. The encoded sequence is
It is generated in the form of a bit stream according to a predetermined format. The encoded sequence can be generated in a format known to those skilled in the art.

FIG. 3 is a block diagram of the decoding device 120 shown in FIG.
1 shows a configuration example of a decoding device 120A as an example.

The decoding device 120A according to the present embodiment receives an encoded sequence and generates an output spectrum sequence. The coded sequence includes a plurality of first codes and at least one second signal. Each of the plurality of first codes is a code generated to represent the average amplitude of the frequency spectrum sequence for one of the plurality of frequency bands. As used herein, the term “first code” refers to a code generated to represent the average amplitude of a frequency spectrum sequence for one of a plurality of frequency bands, and the term “second code” , A code obtained by encoding a frequency spectrum sequence corresponding to the average amplitude represented by the first code.

The coded sequence received by the decoding device 120A is, for example, a coded sequence generated by the coded sequence generating unit 240A of the above-described coding device 110A. Decryption device 1
The output spectrum sequence generated by 20A is converted into a decoded audio signal, which is a time signal, by frequency-time conversion section 30 (see FIG. 1).

The decoding device 120A uses the first code and the second code
Sequence analysis unit 31 for analyzing a coded sequence including a code
0A and a band gain dequantizer 320A that dequantizes the first code to generate an average amplitude of the frequency spectrum sequence.
And in the frequency band of at least one second code,
A coding band notification unit 330A that notifies whether there is a frequency band corresponding to the frequency band of the first code;
A spectrum dequantizer 340A that generates a frequency spectrum sequence obtained by inversely quantizing the code of, a noise spectrum sequence generator 350A that generates a noise spectrum sequence, and a frequency spectrum sequence dequantized by the spectrum dequantizer 340A And noise spectrum sequence generator 350A
Amplifying section 360A that amplifies the noise spectrum sequence generated by the above, and a spectrum combining section 365A that combines the amplified frequency spectrum sequence and the amplified noise spectrum sequence output from amplification section 360A. The amplification unit includes a noise spectrum sequence amplification unit 362A that amplifies the noise spectrum sequence and a spectrum amplification unit 364A that amplifies the frequency spectrum sequence.

The coded sequence analyzer 310A receives the coded sequence and analyzes the received coded sequence.

The band gain dequantization unit 320A performs the following based on the code analyzed by the encoded sequence analysis unit 310A.
A quantized decoded average amplitude qrms for each scale factor band is generated. Quantized average amplitude qrm
s is calculated using the above (Equation 3).

The coded sequence analyzer 310A outputs the first code to the band gain dequantizer 320A. The coded sequence analysis unit 310A also transmits to the coding band notification unit 330A information on whether or not there is a frequency band corresponding to the frequency band of the first code in the frequency band of at least one second code. Output. Coding band notification unit 330A
Is in the frequency band of at least one second code,
It is notified whether there is a frequency band corresponding to the frequency band of the first code. Here, it is assumed that the code sequence is set in advance to include a code obtained by encoding a spectrum sequence of N scale factor bands from the low band of the plurality of scale factor bands. However, the present embodiment is not limited to this.

When the coding band notifying unit 330A notifies the spectrum dequantizing unit 340A that there is a frequency band corresponding to the frequency band of the first code in the frequency band of at least one second code. The spectrum dequantizer 340A dequantizes the second code received from the coded sequence analyzer to generate a frequency spectrum sequence. When the encoding of the second code is performed by Huffman encoding, the spectrum inverse quantization unit 340A performs Huffman decoding. When the encoding of the second code is performed by vector quantization, the spectrum inverse quantization unit 340A performs inverse vector quantization. Here, it is assumed that the type obtained by encoding the second code is set in advance. However, the present embodiment is not limited to this, and the coded sequence includes a code indicating a type obtained by coding the second code, and the spectrum inverse quantization unit 340A converts the second code based on the code. The type of decoding may be determined.

The spectrum amplification unit 364A amplifies the frequency spectrum sequence generated by the spectrum dequantization unit 340A using the average amplitude generated by the band gain dequantization unit 320A.

If the average amplitude generated in a certain scale factor band is qrms, and the frequency spectrum sequence generated by spectrum dequantizing section 340A is qsp (i) corresponding to the scale factor band, the spectrum amplifying section 364A output (rsp)
Is given by (Equation 4).

[0061]

(Equation 4) In the frequency band of at least one second code, the first
When the coding band notification unit 330A notifies the noise spectrum sequence generation unit 340A that there is no frequency band corresponding to the frequency band of the code
50A outputs the noise spectrum to amplifier 360A. In this specification, the noise spectrum refers to a spectrum on a frequency axis. Noise spectrum sequence generator 3
50A may use, as a noise spectrum, a spectrum obtained by performing the same time-frequency conversion as that performed by time-frequency conversion section 20 (see FIG. 1) on a white noise signal prepared in advance. The frequency spectrum of the white noise is normalized such that the average amplitude obtained from (Equation 1) to (Equation 3) is 1. Alternatively, the noise spectrum sequence generation unit 350A may hold the value of the noise spectrum in some recording medium in advance and simply output the value.

The noise spectrum amplification unit 362A amplifies the noise spectrum sequence generated by the noise spectrum sequence generation unit 350A using the average amplitude generated by the band gain dequantization unit 320A. The amplification is performed in the same procedure as in (Equation 4).

As described above, when the coded sequence includes the second code, amplifying section 360A performs spectrum dequantizing section 340A.
The frequency spectrum sequence is amplified based on the frequency spectrum sequence generated by the above and the average amplitude generated by the band gain dequantizer 320A.

Further, the amplifying section 360A determines that the encoded sequence is the second
, The noise spectrum sequence generator 35
The noise spectrum sequence is amplified based on the noise spectrum sequence generated by 0A and the average amplitude generated by the band gain dequantizer 320A. The spectrum combining unit 365A combines the amplified noise spectrum sequence and the amplified spectrum sequence to generate an output spectrum sequence.

To summarize the above, at least one second
If there is a frequency band corresponding to the frequency band of the first code in the frequency band of the code
0A notifies the spectrum dequantizer 340A to generate a frequency spectrum sequence decoded by decoding the second code. In this case, spectrum inverse quantization section 340
A shows a frequency spectrum sequence as a spectrum amplification unit 364A.
Output to The spectrum amplifying unit 364A converts the frequency spectrum sequence into a first code and a band gain
The signal is amplified by the average amplitude dequantized by A.

Alternatively, if there is no frequency band corresponding to the frequency band of the first code in the frequency band of the at least one second code, the coding band notification unit 330A sends the noise spectrum sequence generation unit 350A Notify to output a noise spectrum sequence. In this case, the noise spectrum sequence generator 350A outputs the noise spectrum sequence to the noise spectrum amplifier 362A. The noise spectrum amplifying unit 362A amplifies the noise spectrum sequence by the average amplitude of the first code, which is inversely quantized by the band gain inverse quantization unit 320A.

FIG. 4 shows an output spectrum represented by an output spectrum sequence output from decoding apparatus 120A. The vertical axis of FIG. 4 represents the amplitude of the output spectrum, and the horizontal axis represents the frequency.

FIG. 4 shows that the frequency band is divided into a low band side and a high band side.
Although the case where the coded sequence includes the second code corresponding to the scale factor band on the low frequency side is illustrated, the present embodiment is such that the coded sequence includes the second code in this way. It is not limited to the case where the frequency band is continuous from the low band. The output spectrum represented by the output spectrum sequence output from the amplification unit 360A is converted by the frequency-time conversion unit 30 (see FIG. 1) into a decoded audio signal that is a time signal sequence.

In the above description, a case has been described where the scale factor band to be quantized and encoded by the encoding device 110A and the scale factor band to be decoded by the decoding device 120A are preset. However, the present embodiment is not limited to this. In the present embodiment, the scale factor band quantized and encoded by encoding apparatus 110A may be determined based on the average amplitude or the information amount of the encoded spectrum sequence. Also, the decryption device 120
The scale factor band that A decodes may be determined by the code included in the coded sequence.

FIG. 2B shows the decoding device 11 shown in FIG.
11 shows a configuration example of a decoding device 110B as an example of “0”.

In coding apparatus 110B, coding band determination section 220B should quantize and code based on the information amount of the coding sequence used by band gain coding section 210B to represent the average amplitude of the scale factor band. The frequency band is determined, and the coded sequence generation unit 240B
The encoding device 110A described with reference to FIG. 2A except that the code representing the frequency band determined by 20B is included in the encoded sequence.
Is the same as Band gain coding section 210B, coding band determination section 220B, spectrum coding section 230B, and coded sequence generation section 240B of coding apparatus 110B include band gain coding section 210A and coding band determination section of coding apparatus 110A. 220A, the spectrum encoding unit 230A, and the encoded sequence generation unit 240A.

In accordance with the amount of information of the coded sequence used by band gain coding section 210B to represent the average amplitude of each scale factor band, coding band determination section 220B determines whether spectrum coding section 230B quantizes the code. Determine the number of scale factor bands to be converted.

For example, if the information amount of the coded sequence used to represent the average amplitude is larger than a certain threshold, the coding band determination unit 220B determines whether the spectrum coding unit 230B quantizes the scale factor band to be coded. Reduce the number. Conversely, when the information amount of the coded sequence used to represent the average amplitude is smaller than a certain threshold, the coding band determination unit 220B determines the number of scale factor bands to be quantized and coded by the spectrum coding unit 230B. Do more.

As described above, the coding band determination unit 220B
Can control the number of scale factor bands to be quantized and encoded based on the encoding result of the band gain encoding unit 210B.

In this case, the coded sequence generation unit 240B includes a coding band determination unit 220B in addition to the average amplitude generated by the band gain coding unit 210B and the coded spectrum sequence generated by the spectrum coding unit 230B.
A coded sequence is generated from a code (third code) representing the scale factor band to be coded and quantized determined by B.

FIG. 2C shows the decoding device 11 shown in FIG.
11 shows a configuration example of a decoding device 110C as an example of 0.

Coding apparatus 110C performs coding band determination section 220C based on the information amount of the coding sequence used by spectrum coding section 230C to represent the coding spectrum sequence.
Determine the frequency band to be quantized and encoded, and the coding sequence generation unit 240C includes a code representing the frequency band determined by the coding band determination unit 220C in the coding sequence, except that
This is the same as the encoding device 110A described with reference to FIG. 2A. Band gain coding section 210C, coding band determination section 220C, spectrum coding section 230C of coding apparatus 110C,
The coded sequence generation unit 240C includes a band gain coding unit 210A, a coding band determination unit 220A of the coding device 110A,
Spectrum encoder 230A, encoded sequence generator 240A
Respectively.

For example, if the size of the coded sequence is set in advance and the coding type of the spectrum coding unit 230C is Huffman coding, the coding band determining unit 2
20C determines that Huffman coding is performed in order from the lower frequency band of the plurality of frequency bands. If Huffman coding cannot be performed entirely due to the restriction on the size of the coded sequence, the coding band determination unit 220C
Is determined not to perform Huffman coding on a frequency band higher than a certain frequency band. Also in this case, encoded sequence generating section 240C outputs average amplitude and spectrum encoded section 230C generated by band gain encoding section 210A.
In addition to the coded spectrum sequence generated by the above, a coded sequence is generated from a code (third code) representing the scale factor band to be quantified and coded, determined by the coding band determination unit 220C.

Alternatively, it is conceivable that the encoding band determining section 220C always determines the frequency band to be quantized and encoded beforehand. In this case, based on the size of the second code obtained by quantizing and encoding the predetermined frequency band by the spectrum encoding unit 230C, the frequency encoding unit 230C first quantizes the frequency band from among the frequency bands not determined to be encoded. The frequency band to be further quantified and encoded may be determined again. At this time, the spectrum encoding unit 230C
The frequency spectrum sequence is quantized and encoded for the re-determined frequency band to generate a second code.

As shown in FIGS. 2B and 2C, the coded sequence may include a third code indicating which scale factor band was coded.

The operation of the decoding apparatus in that case will be described with reference to the decoding apparatus 120A shown in FIG.

The coded sequence analyzer 310A analyzes the third code of the coded sequence. The coding band notification unit 330A decodes which scale factor band is being coded based on the third code analyzed by the coded sequence analysis unit 310A, and based on the decoded result, Notifying unit 340A or noise spectrum sequence generating unit 350A. For example, the coding band notification unit 330
A decodes information such as that the N scale factor bands on the lower frequency side are encoded.

The spectrum dequantizing unit 340A performs the following processing based on the result analyzed by the encoded sequence analyzing unit 330A.
The coding band notification unit 330A decodes the frequency spectrum of the scale factor band determined to be coded. When the encoding of the second code is performed by the Huffman encoding, the spectrum inverse quantization unit 340A performs the Huffman decoding of the second code. If the encoding of the second code is performed by vector quantization, the spectrum inverse quantization unit 340
A inverse vector quantizes the second code. Amplifier 36
0A amplifies the decoded frequency spectrum generated by the spectrum dequantizer 340A using the average amplitude decoded by the band gain dequantizer 320A.

The coded sequence generated by such a method is as follows:
Even with a small amount of information, a wideband decoded audio signal can be obtained. This is because it does not encode the fine structure of the spectrum for all wideband signals, but only encodes the average amplitude for some bands, so that the encoded sequence can be reduced. . However, the decoded audio signal generated in this manner retains the average amplitude of each frequency band of the audio signal input at the time of encoding, so that clear reproduction with less muffled sound in a narrow band sound is obtained. Can be provided.

(Embodiment 2) According to Embodiment 2, at the time of encoding, a time signal sequence of one frame representing an audio signal is divided into a plurality of time signal sequences corresponding to a plurality of time areas, and the divided time Embodiment 2 is different from Embodiment 1 in that the average amplitude of the time signal sequence is encoded for each area. In the second embodiment, the fourth code representing the average amplitude of the time domain time signal sequence obtained by dividing the time signal sequence of one frame into a plurality of time signal sequences corresponding to a plurality of time domains at the time of decoding. Is different from the first embodiment in that

FIG. 5 is a block diagram of the encoding device 110 shown in FIG.
1 shows a configuration example of an encoding device 110D as an example.

The coding apparatus 110D further includes a time domain gain coding section 250D for generating a fourth code representing the average amplitude of the time signal sequence in the time domain.
40D is the same as the encoding device 110A described with reference to FIG. 2A, except that 40D includes the fourth code in the encoded sequence. Band gain coding section 210D, coding band determination section 220D, spectrum coding section 230D, and coding sequence generation section 240D of coding apparatus 110D include band gain coding section 210A and coding band determination section of coding apparatus 110A. 220A, the spectrum encoding unit 230A, and the encoded sequence generation unit 240A.

The audio signal is input to the time-frequency converter 20 every predetermined number of samples. The time-frequency conversion unit 20 generates a spectrum on the frequency axis from a signal sequence on the time axis by using a method such as a modified discrete cosine transform (MDCT transform). As described above, the spectrum on the frequency axis is the frequency spectrum of one frame. The frequency spectrum generated by time-frequency conversion section 20 is input to band gain encoding section 210D and encoding band determination section 220D as a frequency spectrum sequence, as described in the first embodiment.

The audio signal is input as an audio discrete signal to the time-domain gain coding section 250D at the same time intervals as the time-frequency conversion section 20. Time domain gain encoding section 250D divides the audio discrete signal into a plurality of temporally continuous time domains.

For example, when the audio signal is represented by 512 samples that are continuous in time, the time-domain gain encoding unit 250D converts the audio signal into in [i] (i =
0, 1, 2,. . . , 511), and is assumed to be divided into four time domains every 128 samples. The data in the 0th time domain is in [i] when i is 0 to 127. In the data of the first time domain, i is from 128 to 25
5 is in [i]. The data in the second time domain is in [i] when i is 256 to 383.
The data in the third time domain is in [i] when i is 384 to 511. Time domain gain encoding section 250
D calculates the average amplitude of each time domain using, for example, (Equation 5).

[0091]

(Equation 5) Here, j is a time domain number, and g [j] is a j-th number.
It is the average amplitude of the th time domain. Next, an average amplitude ratio of each time domain is calculated from the calculated average amplitude of each time domain. For example, if the average amplitude of the time domain having the maximum value among the four time domains is normalized to be 16,
The average amplitude ratio in each time domain is represented by 4 bits. For example, it is calculated as in (Equation 6).

[0092]

(Equation 6) Here, rg (j) is the quantized average amplitude in the j-th time domain, and gmax is the maximum value of g (j). The time domain gain encoding unit 250D calculates the calculated rg
(J) is encoded and sent to the encoded sequence generation unit 240D. In the above description, rg (j) is normalized by 16 so as to be quantized by 4 bits when deriving rg (j).
It does not have to be a bit. For example, the average amplitude ratio in each time domain may be quantized by one bit. As described above, by deriving the average amplitude ratio in each time domain, the average amplitude in the time domain can be represented by a predetermined amount of information.

Further, in the above description, the average amplitude ratio in each time domain is derived, but the present embodiment is not limited to this. A value obtained by simply encoding the average amplitude in each time domain may be transmitted to the encoded sequence generation unit 240D.

FIG. 6 is a block diagram of the decoding device 120 shown in FIG.
1 shows a configuration example of a decoding device 120B as an example.

The decoding apparatus 120B has a time-domain gain decoding section 370B and is similar to the decoding apparatus 12 shown in FIG.
Different from 0A. Coded sequence analysis section 310B and band gain dequantization section 320 of decoding apparatus 120B shown in FIG.
B, the coding band notifying unit 330B, the spectrum dequantizing unit 340B, the noise spectrum sequence generating unit 350B, the amplifying unit 360B, and the spectrum synthesizing unit 365B are each a coded sequence analyzing unit of the decoding device 120A shown in FIG. 31
0A, a band gain dequantizer 320A, a coding band notifier 330A, a spectrum dequantizer 340A, a noise spectrum sequence generator 350A, an amplifier 360A, and a spectrum synthesizer 365A.

In the decoding device 120B, the coded sequence analyzer 330B receives the coded sequence including the fourth code representing the average amplitude of the time signal sequence in the time domain, and analyzes the coded sequence. Time domain gain decoding section 370B decodes the average amplitude of the time domain time signal sequence from the fourth code analyzed by coded sequence analysis section 330B. The average amplitude of the time domain time signal sequence decoded from the fourth code is sent to noise spectrum sequence generation section 350B. The noise spectrum sequence generation unit 350B includes a noise spectrum sequence generation unit 350B configured to convert the noise into a plurality of time domain noise signals based on the fourth code decoded by the time domain gain decoding unit 370B. Generate a spectrum sequence.

Here, when the fourth code is the time domain gain ratio rg (j) representing the ratio of the average amplitude in each time domain, as described above with reference to FIG. 5, the noise spectrum sequence generator The 350B generates an amplified noise signal by multiplying the noise signal by a time domain gain ratio rg (j) according to the group length. For example, this corresponds to generating an amplified noise signal as shown in (Equation 7).

[0098]

(Equation 7) n (i) is a noise signal and an (i) is an amplified noise signal. The noise spectrum sequence generation unit 350B includes:
By performing time-frequency conversion on the amplified noise signal an (i) in the same manner as the time-frequency conversion unit 20, a noise spectrum is generated and output to the amplification unit 360B. Subsequent processing is the same as in the first embodiment. Further, the noise spectrum sequence generating unit 350B may hold the value of the noise spectrum in some recording means in advance and simply output it when needed.

The coded sequence generated by such a method is as follows.
Even with a small amount of information, a wideband decoded audio signal can be obtained. This is because it does not encode the fine structure of the spectrum for all wideband signals, but only encodes the average amplitude for some bands, so that the encoded sequence can be reduced. . However, since the decoded audio signal obtained in this way holds the average amplitude of each frequency band of the audio signal input at the time of encoding,
It is possible to provide a clear reproduced sound with less muffled feeling in a narrow band sound. By decoding the temporal average amplitude, it is possible to provide crisp reproduction.

(Embodiment 3) Embodiment 3 is different from Embodiment 1 in that, at the time of encoding, a frequency band that is quantized and not encoded is divided into sub-bands and an average amplitude of the sub-band is generated. different. Also, in the third embodiment, at the time of decoding,
Embodiment 4 is different from Embodiment 1 in that a fifth code representing the average amplitude of a frequency spectrum sequence of a sub-band obtained by dividing a frequency band to be quantized and not encoded is decoded.

FIG. 7 is a block diagram of the encoding apparatus 110 shown in FIG.
1 shows a configuration example of an encoding device 110E as an example. Encoding device 110E is different from encoding device 110A shown in FIG. 2A in having subband encoding section 260E. Band gain encoding section 210 of encoding apparatus 110E
E, coding band determination unit 220E, spectrum coding unit 2
30E and the coded sequence generation unit 240E correspond to the band gain coding unit 210A, the coded band determination unit 220A, the spectrum coding unit 230A, and the coded sequence generation unit 240 of the coding device 110A, respectively.

The frequency spectrum sequence of the scale factor band determined by the coding band determination unit 220E to be quantized and not to be coded is converted into a sub-band gain coding unit 26.
Input to 0E. Sub-band gain encoding section 260E
Selects all or a part of the frequency spectrum sequence of the frequency band (scale factor band) determined not to be encoded. In this specification, the selected frequency band is referred to as a sub-band gain coding application band.

The sub-band gain coding applicable band may be changed according to the amount of information used for coding by spectrum coding section 230E. For example, when the amount of information encoded by the spectrum encoding unit 230E is larger than a certain threshold, the sub-band gain encoding unit 260E reduces the sub-band gain encoding applicable band. Conversely, when the amount of information encoded by the spectrum encoding 230E is smaller than a certain threshold, the sub-band gain encoding unit 2
60E enlarges the sub-band gain coding application band.

At least one frequency spectrum within the sub-band gain coding application band is divided into sub-bands. At this time, the sub-band may include two or more frequency bands.

In the following, as a specific example, the frequency spectrum within one sub-band gain coding applicable band has 16 data, and this frequency spectrum is divided into the lowest frequency component in the frequency band. Are arranged in order from the frequency spectrum corresponding to the highest frequency component to the frequency spectrum corresponding to the highest frequency component.
It is assumed that the data is divided into three sub-bands each having six or five data.

FIG. 9 schematically shows a frequency spectrum of a sub-band according to the present embodiment. In FIG. 9, sub-band 0 is the sub-band of the lowest frequency spectrum, sub-band 1 is the sub-band of the intermediate frequency spectrum,
Sub-band 2 is the sub-band of the highest frequency spectrum. In each sub-band, the average amplitude of each sub-band is calculated. For example, the average amplitude of each sub-band is calculated using (Equation 8).

[0107]

(Equation 8) Here, a certain sub-band gain coding applicable band is ssp.
(J), which consists of subG [i]
Is the calculated average amplitude of the sub-band i. Sub-band gain encoding section 260E encodes the average amplitude of the sub-band based on whether the calculated average amplitude is larger or smaller than a certain threshold, and encodes the encoded result as encoded sequence generating section 10
Send to 4C. Encoding subGsw [i] indicating whether the calculated average amplitude is larger or smaller than a certain threshold is given as, for example, (Equation 9).

[0108]

(Equation 9) Here, Th is a mounting threshold.

FIG. 8 is a block diagram of the decoding device 120 shown in FIG.
15 shows a configuration example of a decoding device 120C as an example of the above. The decoding device 120C includes a sub-band gain encoding unit 380C
This is different from the decoding device 120A shown in FIG.

The coded sequence analyzer 310C, the band gain dequantizer 320C of the decoding device 120C shown in FIG.
Coding band notification unit 330C, spectrum inverse quantization unit 34
0C, noise spectrum sequence generation section 350C, amplification section 36
0C are the decoding devices 120A shown in FIG.
Coded sequence analyzer 310A, band gain dequantizer 32
0A, the coding band notification unit 330A, the spectrum inverse quantization unit 340A, the noise spectrum sequence generation unit 350A, and the amplification unit 360A.

In decoding apparatus 120C, coded sequence analysis section 310C performs coding including a fifth code representing the average amplitude of the frequency spectrum sequence of the sub-band obtained by dividing the frequency band of the frequency spectrum sequence to be quantized and not coded. The sequence is received and the encoded sequence is analyzed. Sub-band gain decoding unit 38
0C decodes the fifth code analyzed by the coding sequence analysis unit 310C into an average amplitude of the frequency spectrum of the sub-band, and generates a noise spectrum sequence of a plurality of sub-bands based on the decoded average amplitude. .

Therefore, sub-band gain decoding section 380C
First decodes the sub-band gain coding applicable band from the frequency bands notified from the coding band notifying unit 330A not to be quantized and encoded, and The average amplitude of the frequency spectrum sequence of the sub-band in the band is decoded. Then, the noise spectrum output from the noise spectrum sequence generation unit 350C is multiplied by the average amplitude of the frequency spectrum sequence of the sub-band, and output from the sub-band gain decoding unit 380C. The output of the sub-band gain decoding unit 380C is expressed by (Equation 1)
0).

[0113]

(Equation 10) Here, nsp (i) is a noise spectrum, and bn
(I) is a frequency spectrum output from sub-band gain decoding section 380C. Sub-band gain decoding unit 380C
Is input to the amplifier 360A, and the subsequent processing is the same as in the first embodiment.

The coded sequence generated by such a method is as follows:
Even with a small amount of information, a wideband decoded audio signal can be obtained. This does not encode the fine structure of the spectrum for all wideband signals, but only encodes the average amplitude for some frequency bands. is there. However, since the decoded audio signal obtained in this way retains the average amplitude of each frequency band of the audio signal input at the time of encoding, clear reproduction with less muffled feeling in a narrow band sound is obtained. Can be provided. Also, by using sub-band gain decoding section 380C, coding band notification section 330
Even in a frequency band notified that there is no frequency spectrum coded by A, a small amount of information is increased, and reproduction closer to an audio signal becomes possible.

[0115]

As described above, according to the coding apparatus and the decoding apparatus of the present invention, it is possible to provide a coded sequence decoded at a low bit rate into a wideband decoded audio signal. .

According to the present invention, of a plurality of frequency bands, a low frequency component is coded using a compression technique such as Huffman coding for a low frequency component, and a high frequency component is coded using a compression technique such as Huffman coding. It is also possible to encode only the information of the average amplitude without encoding the fine structure.
It is possible to minimize the amount of information consumed by coding of high frequency components. However, in the decoding process, a high frequency component is generated by a noise spectrum, so that a reproduced sound with a wide band can be obtained.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of an audio signal conversion system according to the present invention.

2A is a diagram showing a configuration example of an encoding device as an example of the encoding device 110 shown in FIG.

FIG. 2B is a diagram showing a configuration example of an encoding device as an example of the encoding device 110 shown in FIG.

2C is a diagram showing a configuration example of an encoding device as an example of the encoding device 110 shown in FIG.

FIG. 3 is a diagram showing a configuration example of a decoding device as an example of the decoding device 120 shown in FIG.

FIG. 4 is a graph showing an output spectrum represented by an output spectrum sequence output from the decoding device;

FIG. 5 is a diagram showing a configuration example of an encoding device as an example of the encoding device 110 shown in FIG.

FIG. 6 is a diagram showing a configuration example of a decoding device as an example of the decoding device 120 shown in FIG.

7 is a diagram illustrating a configuration example of an encoding device as an example of the encoding device 110 illustrated in FIG.

8 is a diagram showing a configuration example of a decoding device as an example of the decoding device 120 shown in FIG.

FIG. 9 is a graph schematically showing a frequency spectrum of a sub-band according to the third embodiment.

FIG. 10 is a diagram illustrating a configuration example of a conventional encoding device.

[Explanation of symbols]

 Reference Signs List 10 audio signal conversion system 20 time-frequency conversion unit 30 frequency-time conversion unit 40 reproduction unit 100 data processing system 110 coding device 120 decoding device 210A band gain coding unit 220A coding band determination unit 230A spectrum coding unit 240A coding Column generator 250D Time-domain gain encoder 260E Sub-band gain encoder 310A Encoded column analyzer 320A Band gain inverse quantizer 330A Encoding band reproducer 340A Spectrum inverse quantizer 350A Noise spectrum generator 360A Amplification Section 370B time domain gain decoding section 380C sub-band gain decoding section

Continued on the front page F term (reference) 5D045 DA11 DA20 5J064 AA01 BA09 BA13 BB13 BC16 BC18 BC21

Claims (18)

[Claims] A band gain encoding unit that calculates an average amplitude of a frequency spectrum sequence for each of a plurality of frequency bands to generate a first code representing an average amplitude of the frequency spectrum sequence; A coding band determination unit that quantizes a frequency spectrum sequence and determines at least one frequency band to be coded among the bands; and a frequency spectrum sequence for each of the at least one frequency band determined by the coding band determination unit. A spectrum encoding unit that generates a second code by quantizing and encoding a code sequence; and a coded sequence generation unit that generates a coded sequence based on the first code and the second code. An encoding device comprising: 2. The coding band determination unit, based on a size of the first code representing an average amplitude of the frequency spectrum sequence, determines whether the frequency spectrum sequence is to be quantized and encoded. Item 2. The encoding device according to Item 1. 3. The coding band determining unit, based on a magnitude of a second code generated by the spectrum coding unit for at least one frequency band determined to be quantized and to be coded, From the frequency bands that have not been determined to be quantized and coded, the frequency band to be further quantized and coded is re-determined.The spectrum coding unit quantizes a frequency spectrum sequence for the re-determined frequency band. The encoding device according to claim 1, wherein the second code is generated by encoding and encoding. 4. The method according to claim 1, wherein the encoding band determination unit quantizes and encodes a frequency band among a plurality of frequency bands into a frequency spectrum sequence, the first code, The encoding device according to claim 1, wherein the encoded sequence generation unit generates the encoded sequence based on a second code. 5. The encoding apparatus according to claim 1, wherein said spectrum encoding section performs Huffman encoding. 6. The encoding device according to claim 1, wherein the spectrum encoding unit performs vector quantization. 7. The encoding apparatus according to claim 1, wherein the spectrum encoding section performs Huffman encoding and vector quantization. 8. A fourth signal representing an average amplitude of the time signal sequence by calculating an average amplitude of each time signal sequence in a plurality of time domains converted into respective frequency spectrum sequences of the plurality of frequency bands. The encoding device according to claim 1, further comprising a time-domain gain encoding unit that generates a code. 9. At least one frequency band is divided into a plurality of sub-bands among frequency bands which are not quantized and determined to be encoded by the encoding band determination unit, among the plurality of frequency bands, and The encoding device according to claim 1, further comprising a sub-band gain encoding unit that generates a fifth code representing an average amplitude of each of the sub-bands. 10. The encoding apparatus according to claim 9, wherein at least one of the plurality of subbands includes the two or more frequency spectrum sequences. 11. A decoding device for decoding a coded sequence including a first code and at least one second code, wherein the first code is one of a plurality of frequency bands. , And each of the at least one second code quantizes and encodes a frequency spectrum sequence for one of the plurality of frequency bands. A coded sequence analysis unit that analyzes the coded sequence and detects the first code and the at least one second code, and is detected by the coded sequence analysis unit. A band gain inverse quantization unit that inversely quantizes the obtained first code to an average amplitude of the frequency spectrum sequence; and a frequency band of the at least one second code,
A coding band notification unit that notifies whether there is a frequency band corresponding to the frequency band of the first code; and in the frequency band of the at least one second code,
A spectrum inverse quantization unit that inversely quantizes and decodes the second code into the frequency spectrum sequence based on the notification of the encoding band notification unit that there is a frequency band corresponding to the frequency band of the first code. And in the frequency band of the at least one second code,
A noise spectrum sequence generating unit that generates a noise spectrum sequence based on the notification of the coding band notifying unit that there is no frequency band corresponding to the frequency band of the first code, the frequency spectrum sequence or the noise spectrum sequence Amplifying unit that amplifies based on the average amplitude,
Decryption device. 12. The encoded sequence further includes a third code indicating which frequency band among a plurality of frequency bands has been quantized and encoded into a frequency spectrum sequence, and the encoded band notifying unit includes 3 is decoded, and based on the decoded third code, is there a frequency band corresponding to the frequency band of the first code in the frequency band of the at least one second code? Notify whether or not,
The decoding device according to claim 11. 13. The decoding apparatus according to claim 11, wherein said spectrum inverse quantization unit performs Huffman decoding. 14. The decoding device according to claim 11, wherein said spectrum inverse quantization unit performs vector inverse quantization. 15. The spectrum dequantizer performs Huffman decoding and vector dequantization.
3. The decoding device according to claim 1. 16. The encoding sequence according to claim 4, wherein the encoded sequence represents an average amplitude of each time signal sequence in a plurality of time domains to be converted into respective frequency spectrum sequences in the plurality of frequency bands.
The decoding device according to claim 11, further comprising: a time-domain gain decoding unit configured to further decode the fourth code into an average amplitude of the time signal sequence. 17. The noise spectrum sequence generator, wherein the noise spectrum sequence generator is configured to convert the noise spectrum sequence into noise signals in the plurality of time domains based on a fourth code decoded by the time domain gain decoder. 17. The decoding device according to claim 16, wherein 18. The coding sequence according to claim 5, wherein a fifth code representing an average amplitude of each of a plurality of sub-bands obtained by dividing at least one frequency band among frequency bands not dequantized by the spectrum dequantizer is used. Further comprising: decoding the fifth code to an average amplitude of the sub-band;
The decoding device according to claim 11, further comprising: a sub-band decoding unit configured to generate a noise spectrum sequence of the plurality of sub-bands based on the decoded average amplitude.