CN105047201A - Broadband excitation signal synthesis method based on segmented expansion - Google Patents

Abstract

The present invention provides a broadband excitation signal synthesis method based on segment expansion. The present invention combines the two methods of narrowband excitation spectrum translation and white noise excitation, passes the narrowband excitation signal through a low-pass filter, and outputs the first frequency band excitation signal. The excitation signal of the first frequency band is shifted and high-pass filtered to obtain the excitation signal of the second frequency band. The excitation signal of the third frequency band is obtained by multiplying the energy-adjusted white noise by the gain factor, and finally the excitation signal of the first frequency band and the excitation signal of the second frequency band are obtained. The excitation signal and the third frequency band excitation signal are synthesized into a complete broadband excitation signal. This method can better preserve the harmonic structure of the excitation signal and does not introduce too many artificial harmonics. The algorithm itself has low computational complexity and is easy to implement. , which improves the performance of the existing bandwidth extension technology and improves the communication quality of the narrowband voice.

Description

A Broadband Excitation Signal Synthesis Method Based on Segment Expansion

technical field

The present invention relates to the technical field of speech signal processing, and more specifically, to a method for synthesizing broadband excitation signals based on segment expansion.

Background technique

Due to the limitation of its transmission bandwidth, the traditional telephone network can only communicate with narrowband voice. Although the narrowband voice can meet the basic communication requirements, the communication quality is greatly reduced. Due to economic reasons, it is impossible to realize real broadband transmission in a short period of time on the old telephone network. The emergence of artificial voice bandwidth expansion technology solves this problem. This technology does some processing at the receiving end of the telephone network, and artificially adds some missing frequency band components to the received narrowband voice signal, so that the receiving end outputs broadband with higher auditory quality. Speech signal, after the speech linear prediction model was proposed, most of the artificial speech bandwidth extension technologies are designed and implemented based on this model (also known as the source-filter model). As shown in Figure 1, this model converts the speech bandwidth The expansion is carried out in two independent steps: spectral envelope expansion and excitation signal expansion.

The expanded broadband excitation signal is finally synthesized and output as a broadband speech signal through a synthesis filter (filter coefficients are parameters calculated during the spectrum envelope expansion process). Therefore, excitation signal extension is an essential step in bandwidth extension technology. At present, there are three main excitation signal expansion algorithms: narrowband excitation spectrum folding/spectrum translation, white noise excitation and harmonic excitation. Although these three methods are relatively mature at present, their disadvantages lie in that the appropriate harmonic structure must be preserved in the synthesized excitation signal, and too many artificial harmonics must not be introduced, because too many artificial harmonics will affect the The original harmonic structure will cause damage; the computational complexity of the algorithm should not be too high, otherwise the algorithm will be difficult to implement.

Contents of the invention

The invention provides a method for synthesizing broadband excitation signals based on segment expansion, which has low computational complexity and is easy to implement.

In order to achieve the above-mentioned technical effect, the technical scheme of the present invention is as follows:

A method for synthesizing broadband excitation signals based on segment extension, comprising the following steps:

S1: The narrowband speech signal is subjected to upsampling and frame preprocessing, and then passes through the analysis filter to obtain the narrowband excitation signal u _n ;

S2: Pass the narrow-band excitation signal u _n through a low-pass filter with a cut-off frequency of α1, output the first frequency band excitation signal u _l , perform spectrum shift operation on the first frequency band excitation signal u _l , and pass the spectrum shifted signal through a A high-pass filter with a cutoff frequency of α1 to obtain the second frequency band excitation signal u _m ;

S3: Set up a unit white noise generator to generate white noise signal u _f , calculate the variance of the excitation signal u _l in the first frequency band to obtain the gain factor Adjust the energy of the white noise signal u _f and multiply by the gain factor Then through a high-pass filter with a cut-off frequency of α2, the excitation signal u _h of the third frequency band is obtained;

S4: Combining the excitation signal u _l of the first frequency band, the excitation signal u _m of the second frequency band and the excitation signal u _h of the third frequency band to obtain a broadband excitation signal u.

Further, in the step S3, the frequency band of the white noise signal u _f after energy adjustment is infinitely wide, and a high-pass filter with a cut-off frequency of α2 is used to filter out the spectral components below the white noise α2, and the frequency band range is 6500Hz- A third frequency band excitation signal u _h at 8000 Hz.

Preferably, the value range of the cut-off frequency α1 is: 3000-3500HZ.

Preferably, the value range of the cutoff frequency α2 is: 6000-6500HZ.

Further, in the step S4, the first frequency band excitation signal u _l , the second frequency band excitation signal u _m and the third frequency band excitation signal u _h are all excitation signals with limited frequency bands and their respective frequency band components do not overlap each other, and the three excitation signals The signals are added in the time domain to obtain a broadband excitation signal u with a complete frequency band.

Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

The present invention combines two methods of narrow-band excitation spectrum translation and white noise excitation, passes the narrow-band excitation signal through a low-pass filter, outputs the first frequency band excitation signal, and obtains the second frequency band after spectral translation and high-pass filtering of the first frequency band excitation signal The excitation signal is obtained as the third frequency band excitation signal by multiplying the energy-adjusted white noise by the gain factor, and finally the first frequency band excitation signal, the second frequency band excitation signal and the third frequency band excitation signal are synthesized into a complete broadband excitation signal, This method can better preserve the harmonic structure of the excitation signal without introducing too many artificial harmonics. The algorithm itself has low computational complexity and is easy to implement. It improves the performance of the existing bandwidth expansion technology and improves the performance of narrowband speech. communication quality.

Description of drawings

Fig. 1 is a schematic diagram of a bandwidth extension method based on a source filter model;

Fig. 2 is a schematic diagram of an excitation signal synthesis method based on segment expansion;

Fig. 3 is the structural representation of spectrum translation method;

Fig. 4 is a comparative diagram of wideband speech, narrowband speech, waveform diagram and spectrogram of speech synthesized by the present invention.

Detailed ways

The accompanying drawings are for illustrative purposes only and cannot be construed as limiting the patent;

In order to better illustrate this embodiment, some parts in the drawings will be omitted, enlarged or reduced, and do not represent the size of the actual product;

For those skilled in the art, it is understandable that some well-known structures and descriptions thereof may be omitted in the drawings.

The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Example 1

As shown in Figure 2, a method for synthesizing broadband excitation signals based on segment expansion includes the following steps:

S1: The narrowband speech signal is subjected to upsampling and frame preprocessing, and then passes through the analysis filter to obtain the narrowband excitation signal u _n ;

S2: Pass the narrow-band excitation signal u _n through a low-pass filter with a cut-off frequency of α1, output the first frequency band excitation signal u _l , perform spectrum shift operation on the first frequency band excitation signal u _l , and pass the spectrum shifted signal through a A high-pass filter with a cutoff frequency of α1 to obtain the excitation signal u _m of the second frequency band; wherein, the value range of the cutoff frequency of α1 is: 3000-3500HZ;

S3: Set up a unit white noise generator to generate white noise signal u _f , calculate the variance of the excitation signal u _l in the first frequency band to obtain the gain factor Adjust the energy of the white noise signal u _f and multiply by the gain factor Then through a high-pass filter with a cut-off frequency of α2, the excitation signal u _h of the third frequency band is obtained; the value range of α2 is: 6000-6500HZ;

S4: Combining the excitation signal u _l of the first frequency band, the excitation signal u _m of the second frequency band and the excitation signal u _h of the third frequency band to obtain a broadband excitation signal u.

In this embodiment, first the narrowband speech signal is subjected to preprocessing such as upsampling (sampling rate 8000Hz), framing, etc., and the purpose of upsampling the narrowband signal is to increase the frequency band range of its signal for subsequent frequency band component expansion. The preprocessed narrow-band signal passes through the analysis filter to obtain the narrow-band excitation signal. The filter coefficient is determined by the parameters calculated by the speech linear prediction analysis. The narrow-band excitation signal is used as the input signal for the expansion of the excitation signal, which is a prerequisite for the expansion of the excitation signal. , after the narrow-band excitation signal is up-sampled, the sampling frequency is changed from 8000 Hz to 16000 Hz, and the signal passes through a low-pass filter to obtain a low-frequency band excitation signal. The parameters of the corresponding low-pass filter are as follows:

Sampling frequency: 16000Hz

Passband cut-off frequency: 3400Hz Passband attenuation: 1dB

Stopband cut-off frequency: 3450Hz Stopband attenuation: 60dB

The obtained first frequency band excitation signal u _l is a low frequency band excitation signal, and its frequency range is 0-3400Hz.

Since the narrowband signal is pre-processed by upsampling before obtaining the narrowband excitation signal, there will be redundant frequency band components in the spectrum of the narrowband excitation signal, so a low-pass filter with a cutoff frequency of 3400Hz must be used to filter out the low frequency of the narrowband excitation signal Out-of-band redundant components, the low frequency band excitation signal u _l with a frequency range of 0-3400Hz is obtained; the low frequency band excitation signal u _l outputs the second frequency band excitation signal u _m through the spectrum translation operation module, u _m is the middle frequency band excitation signal, and the spectrum The specific implementation steps of the translation operation module are shown in Figure 3: the low-frequency band excitation u _l is multiplied by a cosine function with a frequency of 3100 Hz and an amplitude of 2, and then passes through a high-pass filter with a cutoff frequency of 3400 Hz, and the output frequency range is 3400 Hz -6500Hz mid-frequency band excitation signal u _m ; from the modulation characteristics of Fourier transform, multiplying the signal by a cosine function with an angular frequency of ω _M is equivalent to performing a frequency band of the signal with a length of ω in the frequency domain The translation of _M and the time-domain expressions of spectral translation are shown in formula (1):

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; c</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

in, represents the k-th data point of the _m -th frame of the extended mid-band excitation signal um, similarly, Represents the kth data point of the mth frame of the low frequency band excitation signal u _l , in this method, ω _M is selected as 2π·3100(rad/s).

The frequency domain expression corresponding to formula (1) is shown in formula (2):

Φ _e (ω)＝Φ _n (ω-ω _M )+Φ _n (ω+ω _M )(2)

Among them, Φ _e (ω) represents the power spectrum of the signal after spectral shifting, and Φ _n (ω) represents the power spectrum of the excitation signal in the low frequency band.

It can be seen from formula (2) that the power spectrum of the signal after spectral shifting has a power spectrum component of Φ _n (ω-ω _M ). Therefore, the spectrally shifted signal needs to pass through a high-pass filter to filter out this redundant component and retain the required frequency band components. The parameters of the filter are set as follows:

Sampling frequency: 16000Hz

Stopband cut-off frequency: 3350Hz Stopband attenuation: 60dB

Passband cut-off frequency: 3400Hz Passband attenuation: 1dB

The power spectrum of the obtained mid-frequency band excitation signal is shown in formula (3):

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

in, Represents the power spectrum of the mid-band excitation signal.

Calculate the variance of the low frequency band excitation signal u _l Set up a unit white noise generator to generate a white noise signal, and calculate the resulting variance The white noise signal is multiplied as a gain factor and passed through a high-pass filter with a cutoff frequency of 6500 Hz to form a third frequency band excitation signal u _h with a frequency range of 6500-8000 Hz, where u _h is a high frequency band excitation signal. The low-frequency excitation signal u _l , the mid-frequency excitation signal u _m and the high-frequency excitation signal u _h are combined to output a complete broadband excitation signal; the standard deviation of the low-frequency excitation signal is calculated for the energy of the high-frequency unit white noise Adjustment, the calculation formula is shown in formula (4):

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}<span\; class="notranslate">\; =</span>\sqrt{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

In the formula, is the kth data point of the low-frequency excitation signal, and μ is the mean value of the low-frequency excitation signal.

The unit white noise signal is generated by a random function, which obeys the standard normal distribution, and its probability density function is shown in formula (5):

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}}\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

growth factor Multiply the unit white noise signal to obtain the energy-adjusted white noise signal. Since the frequency band of white noise is infinite, what is currently needed is a white noise excitation signal with a high frequency band. Therefore, the energy-adjusted white noise signal needs to pass through a high-pass filter to filter out frequency band components other than the high frequency band in the signal. The filter parameters are set as follows:

Sampling frequency: 16000Hz

Stopband cut-off frequency: 6450Hz Stopband attenuation: 60dB

Passband cut-off frequency: 6500Hz Passband attenuation: 1dB

The white noise signal after passing through the high-pass filter is the required high-frequency excitation signal, and its frequency range is 6500-8000Hz.

The low-frequency excitation signal, the mid-frequency excitation signal and the high-frequency excitation signal are added together in the time domain to obtain a complete broadband excitation signal, as shown in formula (6):

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}_{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

In the formula, represents the estimated value of the kth data point of the mth frame of the broadband excitation signal, similarly, and represent the value of the kth data point in the mth frame of the low frequency band excitation signal u _l , the middle frequency band excitation signal u _m and the high frequency band excitation signal u _h respectively. The experimental results are shown in Fig. 4 .

The same or similar reference numbers correspond to the same or similar components;

The positional relationship described in the drawings is only for illustrative purposes and cannot be construed as a limitation to this patent;

Apparently, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the implementation of the present invention. For those of ordinary skill in the art, other changes or changes in different forms can be made on the basis of the above description. It is not necessary and impossible to exhaustively list all the implementation manners here. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (5)

1. A broadband excitation signal synthesis method based on segment extension, is characterized in that, comprises the following steps: S1: The narrowband speech signal is subjected to upsampling and frame preprocessing, and then passes through the analysis filter to obtain the narrowband excitation signal u _n ; S2: Pass the narrowband excitation signal u _n through a low-pass filter with a cut-off frequency of α1, output the first frequency band excitation signal u _l , perform spectrum shift operation on the first frequency band excitation signal u _l , and pass the spectrum shifted signal through a A high-pass filter with a cutoff frequency of α1 to obtain the second frequency band excitation signal u _m ; S3: Set up a unit white noise generator to generate white noise signal u _f , calculate the variance of the excitation signal u _l in the first frequency band to obtain the gain factor Adjust the energy of the white noise signal u _f and multiply by the gain factor Then through a high-pass filter with a cut-off frequency of α2, the excitation signal u _h of the third frequency band is obtained; S4: Combining the excitation signal u _l of the first frequency band, the excitation signal u _m of the second frequency band and the excitation signal u _h of the third frequency band to obtain a broadband excitation signal u. 2. the broadband excitation signal synthesis method based on segment extension according to claim 1, is characterized in that, in the described step S3, the frequency band of the white noise signal u _f after energy adjustment is infinitely wide, and a cut-off frequency is α2 The high-pass filter is used to filter out the spectral components below the white noise α2, and the third frequency band excitation signal u _h with a frequency range of 6500Hz-8000Hz is obtained. 3. The broadband excitation signal synthesis method based on segment extension according to claim 1, characterized in that the value range of the cutoff frequency α1 is: 3000-3500HZ. 4. The broadband excitation signal synthesis method based on segment extension according to any one of claims 1-2, characterized in that the value range of the cutoff frequency α2 is: 6000-6500HZ. 5. The broadband excitation signal synthesis method based on segment extension according to claim 1, characterized in that, in the step S4, the first frequency band excitation signal u _l , the second frequency band excitation signal u _m and the third frequency band excitation signal Both u and _h are excitation signals with limited frequency bands and their respective frequency band components do not overlap each other. The three excitation signals are added in the time domain to obtain a broadband excitation signal u with a complete frequency band.