CN101303858B - Method and apparatus for implementing fundamental tone enhancement post-treatment - Google Patents

Abstract

A post-processing implementation method and device for pitch enhancement, the implementation process including: obtaining the gain of the decoded signal, and judging whether the gain exceeds a predetermined threshold, and after determining that the gain exceeds the predetermined threshold, the The decoded signal is output after long-term post-filtering. Therefore, in the present invention, the configuration of the filter coefficients and the judgment of the threshold are relatively simple, so that the post-processing process can be simplified, and at the same time, a better pitch enhancement effect can also be obtained.

Description

Method and device for realizing post-processing of pitch enhancement

technical field

The invention relates to the technical field of audio decoding, in particular to an adaptive post-processing technology for pitch enhancement in the audio decoding process.

Background technique

During the audio decoding process, in order to improve the perceived effect of the decoded speech, post-processing operations need to be performed on the decoded speech. The purpose of the post-processing is to enhance information related to perceptual quality in the synthesized sound signal, that is, reduce or remove interference information that degrades perceptual quality, so as to improve perceptual quality. At present, the technologies adopted in the post-processing process are generally divided into formant post-processing technology and pitch post-processing technology. In pitch post-processing techniques, the frequency response of the filter needs to be harmonically related.

Taking AMR-WB+ (Adaptive Multi-Rate Wideband plus, enhanced adaptive multi-rate wideband) codec as an example, the post-processing method adopted is a frequency-band-selectable pitch enhancement post-processing algorithm. As shown in Figure 1, in this post-processing algorithm, specifically, the decoded synthesized sound signal is divided into two sub-bands, and for the low-frequency band, an adaptive pitch enhancement filter is firstly used to adjust the low-frequency end pitch between harmonics The noise of the two frequency bands is weakened, and then processed by low-pass filtering; for the other frequency band, it is directly filtered by high-pass filter; finally, the signals of the two frequency bands that have been correspondingly processed are summed to obtain the pitch-enhanced Synthesize the sound signal.

In Figure 1, in order to achieve the purpose of pitch enhancement post-processing, two modules, Pitchenhancer (pitch enhancement) and Low-pass filter (low-pass filter), are used in the low-frequency sub-band. in:

The function of the Pitch enhancer module is to weaken the internal harmonic noise (inter-harmonic noise) at the low-frequency end of the decoded signal to an appropriate degree, and then pass the Low-pass filter to filter out spectral tilt and other unwanted frequencies Component; the implementation process of the Pitch enhancer module uses a time-varying linear filter, and the transfer function of the filter is:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; E.</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; z</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

In the formula, the coefficient α is used to control the reduction degree of internal harmonic noise, and T is the pitch period of the decoded signal.

Described Low-pass filter (low-pass filter) module is linear phase FIR (finite impulse response) low-pass filter, and its cut-off frequency is 5 * Fs/256kHz, and wherein, Fs is internal sampling rate. The delay of this filter is 12 samples. In the implementation process, the register needs to be updated by using the signal state processed by the low-pass filter, and the number of register bits to be updated in each subframe is 2×12.

The value of coefficient α in formula (1) is between 0 and 0.5, and its specific value is given by the following formula:

α＝0.5× _gp (2)

Wherein, g _p is the decoded pitch gain value. When the TCX (transform domain code excitation) codec mode is used in AMR-WB+, the value of α is directly set to 0. At this time, the transfer function of the pitch enhancement filter becomes:

H _E (z) = 1(3)

Equation (3) indicates that the pitch enhancement post-processing operation is not performed for the TCX codec mode.

It can be seen that the above post-processing method can eliminate the noise component between the harmonics of the low-frequency end of the decoded speech signal, so that the perceived quality of the decoded synthesized sound is improved.

In the process of implementing the present invention, the inventors found that the post-processing implementation of pitch enhancement in the above-mentioned prior art has at least the following problems:

It can be seen from the above implementation process that in the pitch enhancement post-processing algorithm, it is necessary to perform frequency division operation on the decoded speech signal first, and perform different filtering processes on different sub-frequency bands, resulting in complicated implementation of the corresponding post-processing process. .

Contents of the invention

Embodiments of the present invention provide a pitch enhancement post-processing implementation method and device, so as to simplify the post-processing process and improve the quality of the audio signal obtained by the post-processing.

An embodiment of the present invention provides a method for implementing pitch enhancement post-processing, including: obtaining the gain of the decoded signal, judging whether the gain exceeds a predetermined threshold, and after determining that the gain exceeds the predetermined threshold, performing a decoding The signal is post-filtered.

The embodiment of the present invention also provides a device for implementing pitch enhancement post-processing, including:

a gain evaluation unit, configured to obtain the gain of the decoded signal;

a threshold judging unit, configured to judge whether the gain of the decoded signal determined by the gain evaluating unit exceeds a predetermined threshold;

The adaptive post-filter is configured to perform long-term post-filtering only on decoded signals whose gains of the decoded signals exceed a predetermined threshold according to the judgment result of the threshold judging unit.

It can be seen from the technical solutions provided by the above embodiments of the present invention that the implementation of the process of filter coefficient configuration and threshold judgment in the embodiments of the present invention is relatively simple, and a better pitch enhancement effect can be obtained. At the same time, in the embodiment of the present invention, pitch enhancement processing is performed on the entire decoded speech signal, without frequency division processing, and without performing low-pass filtering and high-pass filtering operations separately, which further reduces the complexity of the processing process.

Description of drawings

FIG. 1 is a schematic diagram of the implementation principle of the post-processing of pitch enhancement adopted in the prior art;

Fig. 2 is a schematic diagram of the processing process of the method provided by the embodiment of the present invention;

Fig. 3 is a schematic structural diagram of a device provided by an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a gain evaluation unit in an embodiment of the present invention.

Detailed ways

In the embodiment of the present invention, the energy characteristics of the decoded signal are fully utilized, and are compared with the decoded pitch gain and pitch cycle value to obtain the pitch information that best reflects the characteristics of the sound, thereby providing the option to use the pitch enhancement post-processing filter A threshold evaluation and decision implementation scheme for better perceptual quality of the decoded signal.

In the embodiment of the present invention, it may specifically be as follows: firstly, obtain the gain of the decoded signal, and then judge whether the gain exceeds a predetermined threshold, and if so, perform long-term post-filtering on the decoded signal and output it, otherwise, directly output the decoded signal. Wherein, the post-filter used for post-filtering the decoded signal may be an all-zero post-filter.

In addition, in the embodiment of the present invention, if the all-zero post-filter is selected as the post-filter, the local adjustment factor λ and the adaptive global gain G1 involved in the corresponding filter function are also given to further improve the audio frequency. Specific parameter values for perceived quality. Certainly, other types of post-filters may also be used in the present invention to perform post-filtering processing.

To facilitate the understanding of the embodiments of the present invention, the cause of the coding noise between pitch harmonics will be described first.

Taking AMR-WB+ coding as an example, the speech coding part adopts CELP (Code-Excited Linear Prediction, Code-Excited Linear Prediction) coding technology. At the encoding end, the input signal is pre-emphasized, and the 16-order linear prediction analysis is performed, and then the pitch synthesis filter is used to encode it. The expression of described pitch synthesis filter is:

$\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Among them, T is the pitch period, and g _p is the pitch gain.

In the theory of speech perception, the formant part of the speech is more important to the auditory perception than the valley part of the speech; therefore, at a lower coding rate, it is usually necessary to sacrifice the performance of the valley region and try to make the formant coding superior. This makes the troughs likely to contain more perceptually encoded noise than the peaks, including troughs between pitch harmonic peaks.

Based on the above-mentioned causes of encoding noise, at the decoding end, a post-processing filter needs to be reasonably set to reduce the encoding noise so as to obtain better perceptual quality.

The specific implementation process of the embodiment of the present invention will be described below with reference to the accompanying drawings.

The specific implementation of the method for implementing pitch enhancement post-processing in the audio decoding process provided by the embodiment of the present invention is shown in Figure 2, and specifically includes the following steps:

Step 1, determining the gain of the received decoded signal according to the decoded signal;

Specifically, it can be: the ratio of the signal amplitude in an adjacent period is:

$\mathrm{<span\; class="notranslate">\; ration</span>}<span\; class="notranslate">\; =</span>\sqrt{\frac{\mathrm{<span\; class="notranslate">\; 0.01</span>}<span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; syn</span>}<span\; class="notranslate">\; \_</span>{\mathrm{<span\; class="notranslate">\; in</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 0.01</span>}<span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; syn</span>}<span\; class="notranslate">\; \_</span>{\mathrm{<span\; class="notranslate">\; in</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

The ratio ration is compared with the gain obtained from decoding the code stream, and a smaller value is taken as the final gain value of the decoded signal.

Step 2, judging whether the gain determined in step 1 exceeds a predetermined threshold, if yes, execute step 3, otherwise execute step 4;

In the embodiment of the present invention, according to the signal energy characteristics of the current pitch period and adjacent pitch periods of the decoded synthesized sound signal, a judgment threshold E _thr is set for when to use the post-processing filter, that is, when the gain value determined in step 1 When E _com is greater than E _thr , the corresponding long-term post-filtering operation is performed, otherwise the long-term post-filtering process is not performed;

Wherein, the judgment process based on the threshold value E _thr mainly considers that the voiced speech frame has a strong periodicity, that is, the gain g _p ′ decoded from the code stream transmitted by the encoding end can reflect the voiced sound of this characteristic. According to a large number of program debugging and observation of parameter changes, it can be seen that: in voiced frames, the value of g _p ′ is relatively large and close to a stable value; in unvoiced frames, g _p ′ is small, And a large part tends to 0; overall, the value of g _p ′ and the ratio of the signal amplitude of the current pitch period to the signal amplitude of the previous pitch period are roughly similar; taking the AMR-WB+ codec as an example, After a large number of experiments, and comparing the pesq (objective voice quality assessment) difference between the decoded signal and the original sound signal after each experiment, E _thr can be selected as 0.6;

It should be noted that, according to different encoding and decoding frameworks, the value of the threshold can be determined according to specific conditions. For example, in other encoding and decoding processes except AMR-WB+ encoding and decoding, the selection range of the threshold can be 0 ~1 between;

Step 3, output the decoded signal (i.e., the pitch synthesis signal obtained by decoding at the decoder) after long-term post-filtering, and perform step 4;

Specifically, an all-zero post-filter can be used as a post-filter to weaken the noise between the pitch harmonics; among them, in order to ensure that the peak of the pitch harmonic is still at the above frequency, the zero point should be added to the valley between the pitch harmonics At the frequency corresponding to the position, that is, at π/T,..., (2T-1)*π/T, therefore, the form of the all-zero post-filter that can be used is:

H(z)＝G ₁ ×(1+λ×z ^-T )(6)

In formula (6), T is the pitch period, G ₁ is the total gain control of the filter, and λ is a local adjustment factor;

In this step, taking the AMR-WB+ codec as an example, the determination of the pitch period T of the all-zero post-filter can adopt the pitch period determination method adopted in the AMR-WB+ codec, such as using a pitch tracking module The output T is taken as the pitch period. In order to avoid the phenomenon of pitch doubling (double pitch), it is also necessary to calculate the normalized autocorrelation value of the two signals whose delay is T/2. If the normalized autocorrelation value is greater than 0.95, T/2 is used as New pitch period value in post-processing to obtain the corresponding pitch period value more accurately and in real time at the low frequency end;

In this step, the value range of λ is usually between 0 and 1, and its value determines the weighting degree between signals separated by a pitch period. Still taking the AMR-WB+ codec as an example, after experiments, the The λ value mentioned above can be selected as 0.1;

In this step, in order to prevent the signal distortion caused by the post-filter attenuating the noise between pitch harmonics, the adaptive gain control method is used to determine the adaptive global gain G ₁ , and the adaptive global gain G 1 is determined accordingly. The process of gain G ₁ is as follows:

Assuming that the input of the post-processing filter at time k is x(n) and the output is y(n), then the transfer function of (6) can be obtained

y(n)=G ₁ ×[x(n)+λ×x(nT)](7)

For voiced sound frames, according to the strong periodicity of voiced sound, the waveforms in adjacent pitch periods can be regarded as slightly different in amplitude, so we can set

x(n-T)≈gain×x(n)(8)

Substituting (8) into (7), we can get

y(n)≈G ₁ ×[1+λ×gain]×x(n)(9)

It can be seen from the above derivation that if no adaptive gain control is performed, the filter will make the output y(n) much larger than the input x(n) while completing the pitch enhancement post-processing that weakens the inter-harmonic noise. The perceptual quality of the final synthesized speech signal will be greatly reduced; so the value of the adaptive global gain G ₁ is selected as:

${\mathrm{<span\; class="notranslate">\; <figure-callout\; id="1"\; label="G"\; filenames="S071A4394520070606E000012.png"\; state="\{\{state\}\}">G</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; \&ap;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; gain</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; )</span>$

In this way, the parameters of the all-zero post-filter can be determined.

Step 4, outputting the pitch synthesis signal obtained at the decoding end.

Specifically, it can be as follows: suppose that in steps 2 and 3, the decoded pitch synthesis signal is synth_in, and the output signal after the pitch long-term post-filtering process is synth_out, then the processing of steps 2 and 3 can be performed as follows formula means:

$\mathrm{<span\; class="notranslate">\; synth</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; out</span>}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; synth</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; in</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; com</span>}}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; thr</span>}}\\ \mathrm{<span\; class="notranslate">\; synth</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; in</span>}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; if</span>}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; com</span>}}<span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}_{\mathrm{<span\; class="notranslate">\; thr</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span>$

In formula (11), h is the impulse response function of adaptive post-filter H (z); And this formula (11) represents, there are two kinds of pitch synthesis signals output in step 4:

(1) One is after the long-term post-filtering process in step 3, and through the pitch synthesis signal of adaptive gain control, to prevent the signal distortion caused by the post-filter while weakening the noise between the pitch harmonics;

(2) The other is the pitch synthesis signal that is directly output without being processed in step 3.

The embodiment of the present invention also provides a device for implementing pitch enhancement post-processing in the audio decoding process, the specific implementation structure of which is shown in Figure 3, and may specifically include the following processing units:

(1) Gain evaluation unit

This unit is used to obtain the gain of the decoded signal;

Moreover, as shown in Figure 4, the unit may specifically include:

The ratio determination unit is used to determine the ratio of the signal amplitudes of adjacent pitch periods, that is, determine the ratio of the signal amplitudes in the previous pitch period to the signal amplitudes in the current pitch period;

The gain determination unit of the decoded signal is used to compare and select the ratio and the gain obtained by decoding for comparison, and take the smaller value of the two as the gain of the decoded signal.

(2) Threshold judgment unit

The unit is used to judge whether the gain of the decoded signal determined by the gain evaluation unit exceeds a predetermined threshold;

If the device is used in the AMR-WB+ decoding process, the predetermined threshold selected by the threshold judging unit may be 0.6.

(3) Adaptive post filter

It is used to perform long-term post-filtering processing only on decoded signals whose gain of the decoded signal exceeds a predetermined threshold according to the judgment result of the threshold judgment unit;

The adaptive post-filter may be an all-zero post-filter, and the function of the all-zero post-filter is: H(z)=G ₁ ×(1+λ×z ^-T ), where G ₁ is the adaptive global gain, λ is the local adjustment factor, and T is the pitch period;

Moreover, if the device is used in the AMR-WB+ decoding process, the all-zero post-filter adopts: the value of λ is 0.1, and the value of the adaptive global gain $G_{}$${}_1\&ap;\frac{1}{1+\mathrm{\&lambda;}\&times;\mathrm{gain}}$ An all-zero post-filter is used to avoid signal distortion caused by the post-filter while attenuating the noise between the fundamental harmonics.

It should be noted that, in the embodiment of the present invention, the post-filter used for pitch enhancement may also use a comb filter (Comb filter). The comb filter takes advantage of the strong periodicity of voiced sounds. In the frequency domain, the comb filter can retain the fundamental frequency of the sound signal and its integer multiple harmonic components, and suppress the non-harmonic components.

Since the gaps between the harmonics are basically dominated by noise, ideally, if the fundamental frequency (pitch period) is known, the noise between the harmonics can be completely filtered out.

In the embodiment of the present invention, the transfer function of the comb filter is:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; L</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; kT</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$

The corresponding time domain expression is:

$\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; L</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; kT</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 13</span>}<span\; class="notranslate">\; )</span>$

Among them, ×(n) is the decoded speech signal, y(n) is the output after comb filter processing; a _k (-L≤k≤L) is the 2L+1 tap coefficients of the comb filter , the coefficient a _k can be adaptive to the change of the speech signal spectrum. In each subframe, the value of a _k can be configured with reference to the gain of the decoded signal obtained above; for the pitch period T, repeated prediction should be avoided.

It can be seen from equation (13) that the output y(n) is a delay-weighted average of the input x(n) to emphasize the periodic component; when the delay coincides with the pitch period, this averaging process makes the periodicity Components are enhanced, while those aperiodic components or other components that differ from the period of the signal are suppressed or completely eliminated.

In summary, in the embodiment of the present invention, in the case of using the FIR filter to perform pitch enhancement post-processing on the decoded sound signal of the full frequency band, the judgment process of the threshold value and the configuration process of the filter coefficients can be compared. It is simple to implement, and the embodiment of the present invention can also adapt to the energy change of the synthesized sound signal at the decoding end in each subframe, so as to obtain a better pitch enhancement effect. For example, based on the AMR-WB+ codec framework, the post-processing process of pitch enhancement can be realized in a relatively simple operation process, which improves the perceived quality of the decoded sound.

Moreover, the implementation scheme provided by the embodiments of the present invention performs pitch enhancement on the voice signal to obtain better perceptual quality, and through subjective and objective tests on a large number of music sequences, it is found that the degree of improvement in the perceptual quality of the music signal is also very high. big.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present invention can easily think of changes or Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (8)

1. A method for implementing post-processing of pitch enhancement, comprising: obtaining the gain of the decoded signal, judging whether the gain exceeds a predetermined threshold, and after determining that the gain exceeds the predetermined threshold, decoding The signal is post-filtered; Wherein, the step of obtaining the gain of the decoded signal specifically includes: determining the ratio of the signal amplitudes of adjacent pitch periods; The ratio is compared with the pitch gain obtained by decoding, and the smaller value of the two is taken as the gain of the decoded signal. 2. The method according to claim 1, wherein the step of performing post-filtering processing on the decoded signal comprises: The decoded signal is post-filtered using an all-zero post-filter, and the function of the all-zero post-filter is: H(z)=G ₁ ×(1+λ×z ^-T ), where G ₁ is the Adapt to the global gain, λ is the local adjustment factor, and T is the pitch period. 3. The method according to claim 2, characterized in that, in the enhanced adaptive multi-speed wideband AMR-WB+ encoding and decoding process, the described lambda value is selected as 0.1, and the adaptive global gain:

Wherein, gain is the gain of decoding the signal in each subframe. 4. The method according to claim 1, 2 or 3, characterized in that, in the AMR-VB+ encoding and decoding process, the predetermined threshold is 0.6. 5. A device for realizing post-processing of pitch enhancement, characterized in that, comprising: a gain evaluation unit, configured to obtain the gain of the decoded signal; a threshold judging unit, configured to judge whether the gain of the decoded signal determined by the gain evaluating unit exceeds a predetermined threshold; An adaptive post-filter, configured to perform post-filtering processing only on decoded signals whose gain of the decoded signal exceeds a predetermined threshold according to the judgment result of the threshold judgment unit; Wherein, the gain evaluation unit specifically includes: a ratio determination unit, configured to determine the ratio of the signal amplitudes of adjacent pitch periods; The gain determination unit of the decoded signal is configured to compare the ratio with the pitch gain obtained by decoding, and take the smaller value of the two as the gain of the decoded signal. 6. The device according to claim 5, wherein the adaptive post-filter is an all-zero post-filter, and the function of the all-zero post-filter is: H(z)=G ₁ ×(1+λ×z ^−T ), where G ₁ is the adaptive global gain, λ is the local adjustment factor, and T is the pitch period. 7. The device according to claim 6, characterized in that, when the device is used in the AMR-VB+ decoding process, the all-zero post-filter adopts: the λ value is 0.1, and the adaptive global gain

The all-zero post-filter of , where gain is the gain of the decoded signal in each subframe. 8. The device according to claim 5, 6 or 7, wherein when the device is used in the AMR-VB+ decoding process, the predetermined threshold selected by the threshold judging unit is 0.6. the

Images