CN103903634B - Activation tone detection and method and device for activation tone detection - Google Patents

Abstract

The invention relates to an active tone detection (VAD) and a method and a device for active tone detection, the method comprising: obtaining a sub-band signal and a frequency spectrum amplitude of a current frame; calculating values of a frame energy parameter and a spectrum gravity center characteristic parameter of the current frame according to the sub-band signals; calculating to obtain a signal-to-noise ratio parameter of the current frame according to the background noise energy estimated from the previous frame, the frame energy parameter of the current frame and the signal-to-noise ratio sub-band energy; and calculating according to the tonal modification mark, the signal-to-noise ratio parameter, the spectral gravity center characteristic parameter and the frame energy parameter to obtain a VAD judgment result. The method and the device can improve the accuracy of unstable noise (such as office noise) and music detection.

Description

Activation tone detection and method and device for activation tone detection

technical field

The present invention relates to an activation tone detection (VAD) and a method for activation tone detection (including background noise detection, tonal signal detection, correction of the current frame activation tone retention frame number in VAD judgment, signal-to-noise ratio threshold in VAD judgment adjustment methods) and devices.

Background technique

In a normal voice call, the user is sometimes talking and sometimes listening. At this time, there will be an inactive tone phase during the call. Under normal circumstances, the total non-voice activation phase of both parties in the call exceeds 50% of the total voice coding time of both parties in the call. %. During the inactive tone phase, there is only background noise, which usually does not have any useful information. Taking advantage of this fact, in the process of speech and audio signal processing, the active tone and the inactive tone are detected by the activation tone detection (VAD) algorithm, and different methods are used to process them respectively. Many modern speech coding standards, such as AMR and AMR-WB, support the VAD function. In terms of efficiency, the VAD of these encoders does not perform well in all typical background noises. Especially under non-stationary noise, the VAD efficiency of these encoders is low. For music signals, these VADs sometimes have false detections, resulting in significant quality degradation of the corresponding processing algorithms.

Contents of the invention

The technical problem to be solved by the present invention is to provide an activation tone detection (VAD) and a method for activation tone detection (including background noise detection, tonal signal detection, correction of current activation tone retention frame number in VAD judgment, VAD judgment Methods such as the adjustment of the signal-to-noise ratio threshold) and devices, to improve the accuracy of VAD detection.

In order to solve the above-mentioned technical problems, the present invention provides a kind of activation tone detection (VAD) method, and this method comprises:

Obtain the subband signal and spectrum amplitude of the current frame;

The values of the frame energy parameter, the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter of the current frame are calculated according to the sub-band signal; the values of the spectral flatness characteristic parameter and the tonality characteristic parameter are obtained according to the spectral amplitude calculation;

calculating the signal-to-noise ratio parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame and the signal-to-noise ratio sub-band energy;

According to the frame energy parameters of the current frame, the characteristic parameters of the spectral center of gravity, the characteristic parameters of the temporal stability, the characteristic parameters of the spectral flatness, and the characteristic parameters of the tonality, the tonality flag of the current frame is calculated;

The VAD judgment result is calculated according to the tonality mark, the signal-to-noise ratio parameter, the characteristic parameter of the spectral center of gravity, and the frame energy parameter.

In order to solve the above-mentioned technical problems, the present invention provides a kind of activation tone detection (VAD) device, and this device comprises:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

The characteristic parameter acquisition unit is used to calculate the frame energy parameter, spectral center of gravity characteristic parameter and time domain stability characteristic parameter value of the current frame according to the subband signal; obtain the spectral flatness characteristic parameter and tonality characteristic parameter according to the spectral amplitude calculation value;

A sign calculation unit, used to calculate the tonality sign of the current frame according to the current frame frame energy parameter, spectral center of gravity characteristic parameter, temporal stability characteristic parameter, spectral flatness characteristic parameter, and tonality characteristic parameter;

A signal-to-noise ratio calculation unit, which is used to calculate the signal-to-noise ratio parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame, and the signal-to-noise ratio sub-band energy;

The VAD judgment unit is used to calculate and obtain the VAD judgment result according to the tonality flag, the signal-to-noise ratio parameter, the characteristic parameter of the center of gravity of the spectrum, and the frame energy parameter.

In order to solve the above technical problems, the present invention provides a background noise detection method, the method comprising:

Obtain the subband signal and spectrum amplitude of the current frame;

The values of the frame energy parameter, the spectral center of gravity characteristic parameter, and the time domain stability characteristic parameter calculated according to the sub-band signal are obtained, and the values of the spectral flatness characteristic parameter and the tonality characteristic parameter are calculated according to the spectrum amplitude;

Background noise detection is performed according to the characteristic parameters of spectral center of gravity, temporal stability characteristic parameters, spectral flatness characteristic parameters, tonality characteristic parameters, and current frame energy parameters to determine whether the current frame is background noise.

In order to solve the above technical problems, the present invention provides a background noise detection device, which includes:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

The characteristic parameter calculation unit is used to obtain the values of the frame energy parameter, the spectral center of gravity characteristic parameter, and the time domain stability characteristic parameter calculated according to the subband signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude ;

The background noise judging unit is used to detect the background noise according to the spectral center of gravity characteristic parameter, time domain stability characteristic parameter, spectral flatness characteristic parameter, tonality characteristic parameter, and current frame energy parameter, and judge whether the current frame is background noise.

In order to solve the above-mentioned technical problem, the present invention provides a kind of tone signal detection method, and this method comprises:

Obtain the subband signal and spectrum amplitude of the current frame;

Calculate the values of the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter according to the sub-band signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude;

Whether the current frame is a tonality signal is judged according to the characteristic parameter of tonality, the characteristic parameter of time domain stability, the characteristic parameter of spectral flatness, and the characteristic parameter of spectral center of gravity.

In order to solve the above technical problems, the present invention provides a tonality signal detection device, which includes:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

The characteristic parameter calculation unit obtains the values of the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter according to the calculation of the sub-band signal, and obtains the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude calculation;

The tonality signal judging unit is used for judging whether the current frame is a tonality signal according to the tonality characteristic parameter, the temporal stability characteristic parameter, the spectral flatness characteristic parameter, and the spectral center of gravity characteristic parameter.

In order to solve the above-mentioned technical problems, the present invention provides a method for correcting the number of frames of the current frame active tone in the VAD decision, the method comprising:

Calculate the long-term signal-to-noise ratio lt_snr and the average full-band signal-to-noise ratio SNR2_lt_ave;

According to the judgment results of the previous frames, the long-term SNR lt_snr, the average full-band SNR SNR2_lt_ave, the SNR of the current frame, and the VAD judgment result of the current frame, the number of frames to keep the current activation tone is corrected.

In order to solve the above-mentioned technical problems, the present invention provides a correction device for maintaining the number of frames of the current active tone in the VAD decision, the correction device comprising:

A long-term signal-to-noise ratio calculation unit, used to calculate the long-term signal-to-noise ratio lt_snr;

The average full-band SNR calculation unit is used to calculate the average full-band SNR SNR2_lt_ave;

Activation tone keeps the frame number correction unit, is used for according to the decision result of previous several frames, long-term signal-to-noise ratio lt_snr, average full-band signal-to-noise ratio SNR2_lt_ave, the signal-to-noise ratio parameter of current frame and the VAD decision result of current frame, to current Activation sound maintains frame number for correction.

In order to solve the above-mentioned technical problems, the present invention provides a method for adjusting the signal-to-noise ratio threshold in a VAD decision, the adjustment method comprising:

Calculate the spectral center of gravity characteristic parameter of the current frame according to the sub-band signal;

Calculate the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame to obtain the long-term signal-to-noise ratio lt_snr;

Adjust the SNR threshold of the VAD decision according to the characteristic parameters of the spectral center of gravity, the long-term SNR, the number of the previous continuous activation audio frames, and the number of the previous continuous noise frames continuous_noise_num.

In order to solve the above technical problems, the present invention provides a device for adjusting the signal-to-noise ratio threshold in VAD decision, the device for adjusting comprises:

A characteristic parameter acquisition unit, used to calculate the spectral center of gravity characteristic parameter of the current frame according to the subband signal;

The long-term signal-to-noise ratio calculation unit is used to calculate the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame to obtain the long-term signal-to-noise ratio lt_snr;

The signal-to-noise ratio threshold adjustment unit is used to adjust the signal-to-noise ratio threshold of the VAD decision according to the characteristic parameters of the spectral center of gravity, the long-term signal-to-noise ratio, the number of continuous activation audio frames and the number of continuous_noise_num previous continuous noise frames.

The method and device of the invention overcome the shortcomings of the existing VAD algorithm, and improve the accuracy of music detection while improving the detection efficiency of VAD for unstable noise. Therefore, the voice and audio signal processing algorithm using the VAD can obtain better performance.

Description of drawings

Fig. 1 is the schematic diagram of embodiment 1 of activation sound detection method of the present invention;

2 is a schematic diagram of Embodiment 2 of the activation sound detection method of the present invention;

Fig. 3 is the schematic diagram of the process of obtaining VAD judgment result in the embodiment of the present invention 1,2;

Fig. 4 is the module structure schematic diagram of Embodiment 1 of activation sound detection (VAD) device of the present invention;

Fig. 5 is the module structure schematic diagram of Embodiment 2 of activation sound detection (VAD) device of the present invention;

6 is a schematic diagram of the module structure of the VAD decision unit in the VAD device of the present invention;

7 is a schematic diagram of an embodiment of the background noise detection method of the present invention;

8 is a schematic diagram of the module structure of the background noise detection device of the present invention;

FIG. 9 is a schematic diagram of an embodiment of a method for detecting a tone signal in the present invention;

10 is a schematic diagram of the module structure of the tonality signal detection device of the present invention;

Fig. 11 is a schematic diagram of the module structure of the tonality signal judging unit of the tonality signal detection device of the present invention;

FIG. 12 is a schematic diagram of an embodiment of a method for correcting the number of frames maintained by the current active tone in the VAD decision of the present invention;

Fig. 13 is a schematic diagram of the module structure of the correction device for maintaining the frame number of the current active tone in the VAD judgment of the present invention;

FIG. 14 is a schematic diagram of an embodiment of a method for adjusting the SNR threshold in VAD decision of the present invention;

FIG. 15 is a schematic diagram of a specific flow for adjusting the SNR threshold in the present invention;

FIG. 16 is a schematic diagram of the module structure of the device for adjusting the signal-to-noise ratio threshold in the VAD decision of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Apparently, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

Embodiment 1 of the Voice Activity Detection (VAD, Voice Activity Detection) method of the present invention, as shown in Figure 1, the method includes:

Step 101: Obtain the subband signal and spectrum amplitude of the current frame;

In this embodiment, an audio stream with a frame length of 20 ms and a sampling rate of 32 kHz is taken as an example for specific description. Under other conditions of frame length and sampling rate, the method of the present invention is also applicable.

Input the time-domain signal of the current frame into the filter bank, perform sub-band filtering calculation, and obtain the sub-band signal of the filter bank;

In this embodiment, a 40-channel filter bank is used, and the present invention is also applicable to filter banks with other numbers of channels.

Input the time-domain signal of the current frame into the filter bank of 40 channels, perform sub-band filtering calculation, and obtain the filter bank sub-band signal X[k,l] of 40 sub-bands on 16 time samples, 0≤k<40, 0≤l<16, where k is the index of the subband of the filter bank, its value indicates the subband corresponding to the coefficient, l is the time sample index of each subband, and its implementation steps are as follows:

101a: Store the latest 640 audio signal samples in the data buffer.

101b: Move the data in the data buffer by 40 positions, move the earliest 40 sample values out of the data buffer, and store 40 new sample points into positions from 0 to 39.

Multiply the data x in the cache by the window coefficient to get the array z. The calculation formula is as follows:

z[n]=x[n]·W _qmf [n]; 0≤n<640;

Where W _qmf is the filter bank window coefficient.

Use the following pseudo code to calculate a data u of 80 points,

The arrays r and i are calculated using the following equations:

r[n]=u[n]-u[79-n], 0≤n<40

i[n]=u[n]+u[79-n]

101c: Repeat the calculation process of 101b until all the data in this frame is filtered by the filter bank, and the final output result is the filter bank sub-band signal X[k,l].

101d: After completing the above calculation process, the filter bank subband signal X[k, l] of 16 time samples of 40 subbands is obtained, 0≤k<40, 0≤l<16.

Time-frequency transformation is performed on the sub-band signal of the filter bank, and the spectrum amplitude is calculated.

The embodiment of the present invention can be realized by performing time-frequency transformation on all filter bank subbands or part of filter bank subbands and calculating spectrum amplitude. The time-frequency transformation method of the present invention may be DFT, FFT, DCT or DST. This embodiment uses DFT as an example to describe its specific implementation method. The calculation process is as follows:

Perform 16-point DFT transformation on the 16 time sample point data on each filter bank subband indexed from 0 to 9 to further improve the spectral resolution, and calculate the amplitude of each frequency point to obtain the spectral amplitude X _{DFT_AMP} .

The process of calculating the amplitude of each frequency point is as follows:

First, calculate the energy of the array X _DFT [k][j] at each point, the calculation equation is as follows:

X _{DFT_POW} [k, j]=(real(X _DFT [k, j]) ² +(image(X _DFT [k, j]) ² ; 0≤k<10;0≤j<16; where real(X _{DFT_POW} [k,j]), image(X _{DFT_POW} [k,j]) represent the real part and imaginary part of spectral coefficient XD _{FT_POW} [k,j] respectively.

If k is an even number, the following equation is used to calculate the spectrum amplitude at each frequency point:

If k is an odd number, the following equation is used to calculate the spectrum amplitude at each frequency point:

X _{DFT_AMP} is the spectrum amplitude after time-frequency transformation.

Step 102: Calculate and obtain the values of the frame energy parameter and spectral centroid characteristic parameter of the current frame according to the subband signal;

The values of the frame energy parameter, spectral center of gravity characteristic parameter and tonality characteristic parameter can be obtained by prior art methods, preferably, each parameter is obtained by the following method:

The frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy; specifically:

a) Calculate the energy of each filter bank subband according to the filter bank subband signal X[k, l], the calculation equation is as follows:

b) accumulating the energy of some of the filter bank sub-bands that are more sensitive to hearing or all the filter bank sub-bands to obtain the frame energy parameter.

Wherein according to the psychoacoustic model, the human ear is less sensitive to extremely low frequency (such as below 100Hz) and high frequency (such as above 20kHz) sounds, the present invention thinks that according to the filter bank sub-bands arranged from low to high frequency, from the second The penultimate sub-band of the first sub-band is the main filter bank sub-band that is more sensitive to hearing. The frame energy parameter 1 is obtained by accumulating the energy of some or all of the filter bank sub-bands that are more sensitive to hearing. The calculation formula is as follows:

Wherein, e_sb_start is the start subband index, and its value range is [0, 6]. e_sb_end is the end subband index, and its value is greater than 6 and less than the total number of subbands.

The value of frame energy parameter 1 adds the weighted value of the energy of the unused filter bank subband when calculating frame energy parameter 1 partially or completely, obtains frame energy parameter 2, and its calculation equation is as follows:

Among them, e_scale1 and e_scale2 are weighted scale factors, and their value ranges are [0, 1] respectively. _{num_band} is the total number of subbands.

The characteristic parameter of the spectral center of gravity is obtained by calculating the ratio of the sum of the weighted sum of the subband energy of the filter bank to the sum of the direct sum of the subband energy or by smoothing and filtering the values of other characteristic parameters of the spectral centroid.

The characteristic parameters of spectral center of gravity can be realized by the following sub-steps:

a: The sub-band intervals used for the calculation of the characteristic parameters of the spectral center of gravity are divided as follows:

b: Using the spectral centroid characteristic parameter calculation interval division method and the following formula in a, two spectral centroid characteristic parameter values are calculated, which are the spectral centroid characteristic parameter of the first interval and the spectral centroid characteristic parameter of the second interval.

Delta1 and Delta2 are a small offset value respectively, and the value range is (0, 1). where k is the spectral centroid number index.

c: Perform smoothing filtering operation on the characteristic parameter sp_center[0] of the spectral center of gravity in the first interval to obtain the characteristic parameter value of the smoothing spectral center of gravity, that is, the smoothing filtering value of the characteristic parameter value of the spectral center of gravity in the first interval. The calculation process is as follows:

sp_center[2]= _{sp_center} --1 [2]·spc_sm_scale+sp_center[0]·(1-spc_sm_scale)

Among them, spc_sm_scale is the smoothing filter scale factor of spectral center of gravity parameter, sp_center _-1 [2] represents the smoothing spectral center of gravity characteristic parameter value of the previous frame, and its initial value is 1.6.

Step 103: Calculate the SNR parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame, and the SNR subband energy;

The background noise energy of the previous frame can be obtained by existing methods.

If the current frame is the starting frame, the value of the background noise energy of the SNR subband adopts the default initial value. The previous frame SNR sub-band background noise energy estimation is based on the same principle as the current frame SNR sub-band background energy estimation. For the current frame SNR sub-band background energy estimation, refer to step 207 in Embodiment 2 below. Specifically, the signal-to-noise ratio parameter of the current frame can be realized by using an existing signal-to-noise ratio calculation method. Preferably, the following methods are used:

First, the filter bank subband is re-divided into several SNR subbands, and the division index is as follows,

Again, calculate the sub-band average SNR1 according to the energy of each SNR sub-band in the current frame and the background noise energy of each SNR sub-band in the previous frame. The calculation equation is as follows:

Where E _{sb2_bg} is the estimated background noise energy of each SNR sub-band in the previous frame, and num_band is the number of SNR sub-bands. The principle of obtaining the background noise energy of the SNR sub-band of the previous frame is the same as the principle of obtaining the background energy of the SNR sub-band of the current frame. For the process of obtaining the background energy of the SNR sub-band of the current frame, refer to Embodiment 2 below Step 207;

Finally, according to the estimated full-band background noise energy of the previous frame and the frame energy parameters of the current frame, the full-band signal-to-noise ratio SNR2 is calculated:

Where E _{t_bg} is the estimated full-band background noise energy of the previous frame, the principle of obtaining the full-band background noise energy of the previous frame is the same as the principle of obtaining the full-band background noise energy of the current frame, and obtaining the full-band background noise energy of the current frame Refer to step 207 of Embodiment 2 below for the process;

The signal-to-noise ratio parameters in this embodiment include sub-band average signal-to-noise ratio SNR1 and full-band signal-to-noise ratio SNR2. The full-band background noise energy and the background noise energy of each sub-band are collectively referred to as the background noise energy.

Step 104: Calculate and obtain a VAD decision result according to the tonality flag, the signal-to-noise ratio parameter, the characteristic parameter of the center of gravity of the spectrum, and the frame energy parameter.

Example 2

Embodiment 2 of the activation tone detection (VAD) method of the present invention performs polyphase filtering on the input audio signal in sub-frames to obtain the filter bank sub-band signal, and further performs time-frequency conversion on the filter bank sub-band signal, and calculates Spectrum amplitude, signal feature extraction is performed on each filter bank sub-band signal and spectrum amplitude respectively, and each characteristic parameter value is obtained. According to the value of the characteristic parameter, the background noise mark and the tonality mark of the current frame are obtained. According to the current frame energy parameter value and background noise energy calculation, the SNR parameter of the current frame is obtained, and according to the calculated SNR parameter of the current frame, the VAD (Voice Activity Detection, Voice Activity Detection) judgment result and each feature of the previous frame Parameter to determine whether the current frame is an active audio frame. According to the judgment result of the active sound frame, the background noise mark is corrected to obtain a new background noise mark. Whether to update the background noise is judged according to the new background noise identifier. The specific process of VAD detection is as follows:

As shown in Figure 2, the method embodiment 2 includes:

Step 201: Obtain the subband signal and spectrum amplitude of the current frame;

Step 202: Calculate and obtain the values of the current frame energy parameter, spectral center of gravity characteristic parameter, and time-domain stability characteristic parameter according to the sub-band signal; obtain the values of the spectral flatness characteristic parameter and tonality characteristic parameter according to the spectrum amplitude;

The frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy;

The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy;

specifically,

The characteristic parameters of the spectral center of gravity are calculated according to the energy of each filter bank subband, and the characteristic parameter of the spectral center of gravity is obtained by calculating the ratio of the weighted sum of the filter bank subband energy to the direct sum of the subband energy or by other It is obtained by smoothing and filtering the characteristic parameter values of the spectral center of gravity.

The characteristic parameters of spectral center of gravity can be realized by the following sub-steps:

a: The sub-band intervals used for the calculation of the characteristic parameters of the spectral center of gravity are divided as follows:

b: Using the spectral centroid characteristic parameter calculation interval division method and the following formula in a, two spectral centroid characteristic parameter values are calculated, which are the spectral centroid characteristic parameter of the first interval and the spectral centroid characteristic parameter of the second interval.

Delta1 and Delta2 are a small offset value respectively, and the value range is (0, 1). where k is the spectral centroid number index.

c: Perform smoothing filtering operation on the characteristic parameter sp_center[0] of the spectral center of gravity in the first interval to obtain the characteristic parameter value of the smoothing spectral center of gravity, that is, the smoothing filtering value of the characteristic parameter value of the spectral center of gravity in the first interval. The calculation process is as follows:

sp_center[2]=sp_center _-1 [2]·spc_sm_scale+sp_center[0]·(1-spc_sm_scale)

Among them, spc_sm_scale is the smoothing filter scaling factor of spectral center of gravity parameter, sp_center _-1 [2] represents the smoothing spectral center of gravity characteristic parameter value of the previous frame, and its initial value is 1.6.

The time-domain stability characteristic parameter is the expected ratio of the variance of the amplitude superposition value to the square of the amplitude superposition value, or the ratio is multiplied by a coefficient;

specifically,

The time-domain stability characteristic parameters are calculated from the frame energy parameters of the latest several frame signals. In this embodiment, the frame energy parameters of the latest 40-frame signals are used to calculate the time-domain stability characteristic parameters. The specific calculation steps are:

First, calculate the energy amplitude of the latest 40 frames of signals, and the calculation equation is as follows:

Among them, e_offset is an offset value, and its value range is [0, 0.1]

Secondly, the energy amplitudes of the two adjacent frames from the current frame to the previous 40th frame are sequentially added to obtain 20 amplitude superposition values. The specific calculation equation is as follows:

Amp _t2 (n)＝Amp _t1 (-2n)+Amp _t1 (-2n-1); 0≤n＜20;

Wherein, when n=0, Amp _t1 represents the energy amplitude of the current frame, and when n<0, Amp _t1 represents the energy amplitude of n frames before the current frame.

Finally, the time-domain stability characteristic parameter 1td_stable_rate0 is obtained by calculating the ratio of the variance and the average energy of the latest 20 amplitude superposition values. The calculation formula is as follows:

The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of certain spectral amplitudes, or the ratio is multiplied by a coefficient;

Specifically, the spectral amplitude X _{DFT_AMP} is divided into several frequency bands, and the spectral flatness of each frequency band of the current frame is calculated to obtain the characteristic parameter of the spectral flatness of the current frame.

In this embodiment, the spectrum amplitude is divided into 3 frequency bands, and the spectral flatness characteristics of these 3 frequency bands are calculated. The specific implementation steps are as follows:

First, divide X _{DFT_AMP} into 3 frequency bands according to the indexes in the table below.

Secondly, the spectral flatness of each sub-band is calculated respectively to obtain the characteristic parameter of the spectral flatness of the current frame. The calculation equation of each spectral flatness characteristic parameter value of the current frame is as follows:

Finally, smoothing and filtering are performed on the spectral flatness characteristic parameters of the current frame to obtain the final spectral flatness characteristic parameters of the current frame.

sSMR(k)=smr_scale·sSMR _-1 (k)+(1-smr_scale)·SMR(k); 0≤k<3

Among them, smr_scale is the smoothing factor, and its value range is [0.6, 1], and sSMR _-1 (k) is the value of the kth spectral flatness characteristic parameter of the previous frame. .

The tonality feature parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of the two frames of signals before and after, or by continuing to smooth and filter the correlation value.

Specifically, the calculation method of the correlation value of the intra-frame spectral difference coefficients of the two frames of signals before and after is as follows:

The tonality characteristic parameters are calculated according to the spectrum amplitudes, wherein the tonality characteristic parameters can be calculated according to all or part of the spectrum amplitudes.

Its calculation steps are as follows:

a. Partial (not less than 8 spectral coefficients) or all spectral amplitudes are differentially calculated with adjacent spectral amplitudes, and the value of the differential result less than 0 is set to 0 to obtain a set of non-negative spectral differential coefficients.

In this embodiment, frequency point coefficients with position indexes ranging from 3 to 61 are selected as an example to calculate tonality characteristic parameters. The specific process is as follows:

The adjacent spectrum amplitudes from frequency point 3 to frequency point 61 are differentially calculated, and the equation is as follows:

spec_dif[n-3]=X _{DFT_AMP} (n+1)-X _{DFT_AMP} (n); 3≤n<62;

Zero the variables in spec_dif that are less than 0.

b. Calculating the correlation coefficient between the current frame non-negative spectral difference coefficient calculated in step a and the previous frame non-negative spectral difference coefficient to obtain the first tonality characteristic parameter value. The calculation formula is as follows:

Among them, pre_spec_dif is the non-negative spectral difference coefficient of the previous frame.

c, performing a smoothing operation on the first tonality characteristic parameter value to obtain the second tonality characteristic parameter value. The calculation equation is as follows:

tonality_rate2 = tonal_scale tonality_rate2 _-1 + (1-tonal_scale) tonality_rate1

tonal_scale is the smoothing factor of the tonal characteristic parameter, and its value range is [0.1, 1], and tonality_rate2 _-1 is the second tonal characteristic parameter value of the previous frame, and its initial value ranges from [0, 1].

Step 203: Calculate the SNR parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame, and the SNR subband energy;

Step 204: According to the frame energy parameters of the current frame, the characteristic parameters of the spectral center of gravity, the characteristic parameters of the time domain stability, the characteristic parameters of the spectral flatness, and the characteristic parameters of the tonality, calculate and obtain the initial background noise mark and the tonality mark of the current frame;

Step 205: Calculate and obtain the VAD judgment result according to the tonality flag, signal-to-noise ratio parameter, spectral center of gravity characteristic parameter, and frame energy parameter;

Specifically, for a specific implementation method of step 205, refer to the description in conjunction with FIG. 3 below.

Understandably, as long as there is no causal relationship between the parameters of the steps before the VAD decision in step 205, the sequence can be adjusted. For example, the step 204 of obtaining the initial background noise flag and the tonality flag can be before the signal-to-noise ratio calculation step 203.

The initial background noise identifier of the current frame needs to be corrected for the calculation of the SNR parameter of the next frame, so the operation of obtaining the initial background noise identifier of the current frame may also be performed after the VAD decision.

Step 206: Correct the initial background noise identifier according to the decision result of the current frame VAD, the tonality characteristic parameter, the signal-to-noise ratio parameter, the tonality flag, and the time domain stability characteristic parameter;

If the signal-to-noise ratio parameter SNR2 is less than a set threshold value SNR2_redec_thr1, SNR1 is less than SNR1_redec_thr1, the VAD flag vad_f1ag is equal to 0, the tonality characteristic parameter tonality_rate2 is less than tonality_rate2_thr1, the tonality flag tonality_flag is equal to 0 and the time domain stability characteristic parameter lt_stable_rate0 is less than lt_stable_rate0_redec_thr1 (set to 0.1), the background noise flag is assigned a value of 1.

Step 207: Obtain the background noise energy of the current frame according to the correction value of the background noise flag, the frame energy parameter of the current frame, and the full-band background noise energy of the previous frame; the background noise energy of the current frame is used for the next frame signal Noise ratio parameter calculation.

Whether to update the background noise is judged according to the background noise flag. If the background noise flag is 1, the background noise is updated according to the ratio of the estimated full-band background noise energy to the energy of the current frame signal. Background noise energy estimation includes sub-band background noise energy estimation and full-band background noise energy estimation.

a, The subband background noise energy estimation equation is as follows:

E _{sb2_bg} (k) = E _{sb2_bg_pre} (k) · α _{bg_e} + E _{sb2_bg} (k) · (1-α _{bg_e} ); 0≤k<num_sb

Wherein num_sb is the number of frequency domain subbands, E _{sb2_bg_pre} (k) represents the subband background noise energy of the kth SNR subband of the previous frame.

α _{bg_e} is the background noise update factor, and its value is determined by the energy of the full-band background noise of the previous frame and the energy parameters of the current frame. The calculation process is as follows:

If the full-band background background noise energy E _{t_bg} of the last frame is smaller than the frame energy parameter E _t1 of the current frame, the value is 0.96; otherwise, the value is 0.95.

b, Full-band background noise energy estimation:

If the background noise flag of the current frame is 1, then update the background noise energy accumulation value E _{t_sum} and the background noise energy accumulation frame number N _{Et_counter} , the calculation equation is as follows:

E _{t_sum} = E _{t_sum_-1} + E _t1 ;

N _{Et_counter} = N _{Et_counter_-1} +1;

Among them, _{Et_sum_-1} is the background noise energy accumulated value of the previous frame, and N _{Et_counter_-1} is the background noise energy accumulated frame number calculated in the previous frame.

c. The full-band background noise energy is obtained from the ratio of the background noise energy accumulation value E _{t_sum} and the accumulated frame number N _{Et_counter} :

Determine whether N _{Et_counter} is equal to 64, and if N _{Et_counter} is equal to 64, multiply the background noise energy accumulation value E _{t_sum} and the accumulated frame number N _{Et_counter} by 0.75.

d, Adjust the sub-band background noise energy and the accumulated value of the background noise energy according to the tonality flag, the frame energy parameter, and the value of the full-band background noise energy. The calculation process is as follows:

If the tonality flag tonality_flag is equal to 1 and the value of the frame energy parameter E _t1 is smaller than the value of the background noise energy characteristic parameter E _{t_bg} multiplied by a gain coefficient gain,

Then, E _{t_sum} =E _{t_sum} ·gain+delta; E _{sb2_bg} (k)=E _{sb2_bg} (k)·gain+delta;

Among them, the value range of gain is [0.3, 1].

In Embodiment 1 and Embodiment 2, the process of obtaining the VAD judgment result according to the tonality flag, signal-to-noise ratio parameter, spectral center of gravity characteristic parameter, and frame energy parameter calculation includes the following steps as shown in Figure 3:

Step 301: Calculate the long-term signal-to-noise ratio lt_snr through the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame;

See step 307 for the calculation and definition of the average long-duration activation tone signal energy E _fg and the average long-duration background noise energy E _bg . The long-term signal-to-noise ratio lt_snr calculation equation is as follows:

In this formula, the long-term signal-to-noise ratio lt_snr is expressed in logarithm.

Step 302: Calculate the average value of the full-band SNR SNR2 of several recent frames to obtain the average full-band SNR SNR2_lt_ave;

The calculation equation is as follows:

SNR2(n) represents the full-band SNR SNR2 value of the nth frame before the current frame, F_num is the total number of frames for calculating the average value, and its value range is [8, 64].

Step 303: Obtain the signal-to-noise ratio threshold snr_thr of the VAD decision according to the characteristic parameter of the spectral center of gravity, the long-term signal-to-noise ratio lt_snr, the number of continuous active audio frames continuous_speech_num and the number of continuous noise frames before it continuous_noise_num;

The specific implementation steps are as follows:

First, set the initial value of the SNR threshold snr_thr, the range is [0.1, 2], preferably 1.06.

Secondly, the value of the SNR threshold snr_thr is adjusted for the first time according to the characteristic parameter of the spectral center of gravity. The steps are as follows: if the value of the characteristic parameter sp_center[2] of the spectral center of gravity is greater than a set threshold value spc_vad_dec_thr1, then snr_thr is added with a bias value, and the preferred bias value is 0.05; otherwise, if sp_center[1] If it is greater than spc_vad_dec_thr2, add an offset value to snr_thr, and the preferred offset value is 0.10; otherwise, add an offset value to snr_thr, and the preferred offset value is 0.40; among them, the threshold values spc_vad_dec_thr1 and spc_vad_dec_thr2 take values range is [1.2, 2.5]

Again, the value of snr_thr is adjusted twice according to the number of continuous active sound frames continuous_speech_num, the number of continuous noise frames continuous_noise_num, the average full-band SNR SNR2_lt_ave, and the long-term SNR lt_snr. If the number of continuous speech continuous_speech_num is greater than a set threshold value cpn_vad_dec_thr1, then subtract 0.2 from snr_thr; otherwise, if the number of continuous noises in front of continuous_noise_num is greater than a set threshold value cpn_vad_dec_thr2, and SNR2_lt_ave is greater than a bias value plus Multiply the long-term signal-to-noise ratio lt_snr by the coefficient lt_tsnr_scale, then add an offset value to snr_thr, and the preferred offset value is 0.1; otherwise, if continuous_noise_num is greater than a set threshold value cpn_vad_dec_thr3, add an offset value to snr_thr Set value, the preferred offset value is 0.2; otherwise, if continuous_noise_num is greater than a set threshold value cpn_vad_dec_thr4, then an offset value is added to snr_thr, and the preferred offset value is 0.1. Among them, the value range of the threshold values cpn_vad_dec_thr1, cpn_vad_dec_thr2, cpn_vad_dec_thr3, and cpn_vad_dec_thr4 is [2, 500], and the value range of the coefficient lt_tsnr_scale is [0, 2]. Skip this step, directly enter the last step, also can realize the present invention.

Finally, according to the value of the long-term SNR lt_snr, the SNR threshold snr_thr is finally adjusted to obtain the SNR threshold snr_thr of the current frame.

The correction equation is as follows:

snr_thr=snr_thr+(lt_tsnr-thr_offset) thr_scale;

Wherein, thr_offset is an offset value, and its value range is [0.5, 3]; thr_scale is a gain coefficient, and its value range is [0.1, 1].

Step 304: Calculate the initial VAD decision according to the VAD decision threshold snr_thr and the signal-to-noise ratio parameters SNR1 and SNR2 calculated by the current frame;

The calculation process is as follows:

If SNR1 is greater than the decision threshold snr_thr, then judge that the current frame is an active audio frame, and use the value of the VAD sign vad_flag to indicate whether the current frame is an active audio frame. In this embodiment, the value 1 is used to indicate that the current frame is an active audio frame, and 0 indicates that the current frame is an active audio frame. Frames are inactive audio frames. Otherwise, it is determined that the current frame is an inactive audio frame, and the value of the VAD flag Vad_flag is set to 0.

If the SNR2 is greater than a set threshold snr2_thr, it is judged that the current frame is an active audio frame, and the value of the VAD flag vad_flag is set to 1. Among them, the value range of snr2_thr is [1.2, 5.0]

Step 305: Correct the decision result of VAD according to the tonality flag, the average full-band SNR SNR2_lt_ave, the center of gravity of the spectrum, and the long-term SNR lt_snr;

Specific steps are as follows:

If the tonality flag indicates that the current frame is a tonality signal, that is, the tonality_flag is 1, then it is judged that the current frame is an active tone signal, and the vad_flag flag is set to 1.

If the average full-band signal-to-noise ratio SNR2_lt_ave is greater than a set threshold SNR2_lt_ave_t_thr1 plus the long-term signal-to-noise ratio lt_snr multiplied by the coefficient lt_tsnr_tscale, it is judged that the current frame is an active audio frame, and the vad_flag flag is set to 1.

Wherein, the value range of SNR2_lt_ave_thr1 in this embodiment is [1, 4], and the value range of lt_tsnr_tscale is [0.1, 0.6].

If the average full-band SNR SNR2_lt_ave is greater than a set threshold SNR2_lt_ave_t_thr2, and the spectral center of gravity characteristic parameter sp_center[2] is greater than a set threshold sp_center_t_thr1 and the long-term SNR lt_snr is less than a set threshold lt_tsnr_t_thr1, then judge the current The frame is an active audio frame, and the vad_f1ag flag is set to 1. Wherein, the value range of SNR2_lt_ave_t_thr2 is [1.0, 2.5], the value range of sp_center_t_thr1 is [2.0, 4.0], and the value range of lt_tsnr_t_thr1 is [2.5, 5.0].

If SNR2_lt_ave is greater than a set threshold SNR2_lt_ave_t_thr3, and the spectral center of gravity feature parameter sp_center[2] is greater than a set threshold sp_center_t_thr2 and the long-term signal-to-noise ratio lt_snr is less than a set threshold lt_tsnr_t_thr2, then it is judged that the current frame is an active audio frame, The vad_flag flag is set to 1. Wherein, the value range of SNR2_lt_ave_t_thr3 is [0.8, 2.0], the value range of sp_center_t_thr2 is [2.0, 4.0], and the value range of lt_tsnr_t_thr2 is [2.5, 5.0].

If SNR2_lt_ave is greater than a set threshold SNR2_lt_ave_t_thr4, and the spectral center of gravity characteristic parameter sp_center[2] is greater than a set threshold sp_center_t_thr3 and the long-term signal-to-noise ratio lt_snr is less than a set threshold lt_tsnr_t_thr3, then it is judged that the current frame is an active audio frame, The vad_flag flag is set to 1. Wherein, the value range of SNR2_lt_ave_t_thr4 is [0.6, 2.0], the value range of sp_center_t_thr3 is [3.0, 6.0], and the value range of lt_tsnr_t_thr3 is [2.5, 5.0].

Step 306: according to the judgment results of several previous frames, the long-term signal-to-noise ratio lt_snr, the average full-band signal-to-noise ratio SNR2_lt_ave, the signal-to-noise ratio parameter of the current frame and the VAD judgment result of the current frame, modify the number of frames to keep the active tone;

The specific calculation steps are as follows:

The precondition for modifying the number of frames to hold the current active sound is that the active sound flag indicates that the current frame is an active sound frame. If this condition is not met, the value of the number of frames to hold the current active sound num_speech_hangover is not corrected, and directly proceeds to step 307 .

The steps to correct the number of frames for activation sound retention are as follows:

If the previous continuous speech frame number continuous_speech_num is less than a set threshold value continuous_speech_num_thr1, and lt_tsnr is less than a set threshold value lt_tsnr_h_thr1, then the number of currently active tone hold frames num_speech_hangover is equal to the minimum number of continuous active sound frames minus the previous continuous speech frame Number continuous_speech_num. Otherwise, if SNR2_lt_ave is greater than a set threshold value SNR2_lt_ave_thr1, and the number of continuous speech frames continuous_speech_num is greater than a set threshold value continuous_speech_num_thr2, then set the number of active tone hold frames num_speech_hangover according to the size of the long-term signal-to-noise ratio lt_tsnr value. Otherwise, the value of the number of frames num_speech_hangover for the currently active voice will not be modified. Wherein the value of the minimum number of continuous active sound frames in this embodiment is 8, which can be a value between [6, 20].

Specific steps are as follows:

If the long-term signal-to-noise ratio lt_snr is greater than 2.6, the value of num_speech_hangover is 3; otherwise, if the long-term signal-to-noise ratio lt_snr is greater than 1.6, the value of num_speech_hangover is 4; otherwise, the value of num_speech_hangover is 5.

Step 307: Add the active tone hold according to the decision result of the current frame and the number of active tone hold frames num_speech_hangover to obtain the VAD decision result of the current frame.

Its method is:

If the current frame is judged as an inactive tone, that is, the active tone flag is 0, and the number of active tone hold frames num_speech_hangover is greater than 0, add the active tone hold, that is, set the active tone flag to 1, and decrease the value of num_speech_hangover by 1.

Obtain the final VAD decision result of the current frame.

Preferably, after step 304, it also includes calculating the average long-duration activation tone signal energy E _{fg according to the VAD initial decision result;} after step 307, it also includes calculating the average long-duration background noise energy E _bg according to the VAD decision result, and the calculated value is used In the next frame VAD decision.

The specific calculation process of the average long-duration activation tone signal energy E _fg is as follows:

a), if the VAD initial decision result indicates that the current frame is an active sound frame, that is, the value of the VAD flag is 1, and E _t1 is greater than several times of E _bg , which is 6 times in this embodiment, then the average long-term active sound energy accumulation is updated The value fg_energy and the average long-duration activation sound energy accumulation frame number fg_energy_count. The update method is to add E _t1 to fg_energy to get new fg_energy. Add 1 to fg_energy_count to get new fg_energy_count.

b) In order to ensure that the average long-duration activation tone signal energy can reflect the latest activation tone signal energy, if the accumulated frame value of the average long-duration activation tone energy is equal to a certain set value fg_max_frame_num, the accumulated frame number and the accumulated value are multiplied by a Attenuation coefficient attenu_coef1. In this embodiment, the value of fg_max_frame_num is 512, and the value of attenu_coef1 is 0.75.

c), the average long-duration activation tone signal energy is obtained by dividing the average long-duration activation tone energy accumulation value fg_energy by the average long-duration activation tone energy accumulation frame number, and the calculation formula is as follows:

The calculation method of the average long-term background noise energy E _bg is:

Assume that bg_energy_count is the number of accumulated frames of background noise energy, which is used to record how many frames of energy are included in the accumulated value of the latest background noise energy. bg_energy is the cumulative value of the latest background noise energy.

a) If the current frame is determined to be an inactive audio frame, the value of the VAD flag is 0, and the SNR2 is less than 1.0, then update the background noise energy accumulation value bg_energy and the background noise energy accumulation frame number bg_energy_count. The update method is to add the background noise energy accumulation value bg_energy to E _t1 to obtain a new background noise energy accumulation value bg_energy. Add 1 to the background noise energy accumulation frame number bg_energy_count to get the new background noise energy accumulation frame number bg_energy_count.

b) If the background noise energy accumulation frame number bg_energy_count is equal to the maximum count frame number of the average long-term background noise energy calculation, then the accumulation frame number and the accumulation value are multiplied by the attenuation coefficient attenu_coef2 at the same time. Wherein, in this embodiment, the maximum number of counted frames for calculating the average long-term background noise energy is 512, and the attenuation coefficient attenu_coef2 is equal to 0.75.

c), by dividing the background noise energy accumulation value bg_energy by the number of background noise energy accumulation frames to obtain the average long-term background noise energy calculation equation is as follows:

In order to realize the above-mentioned activation tone detection method embodiments 1 and 2, the present invention also provides an activation tone detection (VAD) device embodiment 1, as shown in Figure 4, the device includes:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

A feature parameter acquisition unit, configured to calculate the values of the frame energy parameter and the spectral center of gravity feature parameter of the current frame according to the subband signal;

A signal-to-noise ratio calculation unit, which is used to calculate the signal-to-noise ratio parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame, and the signal-to-noise ratio sub-band energy;

The VAD judgment unit is used to calculate and obtain the VAD judgment result according to the tonality flag, the signal-to-noise ratio parameter, the characteristic parameter of the center of gravity of the spectrum, and the frame energy parameter.

Corresponding to method embodiment 2, the characteristic parameter acquisition unit is also used to calculate the value of the time domain stability characteristic parameter according to the subband signal, and to obtain the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectrum amplitude the value of;

Each characteristic parameter can be obtained by existing methods, or by the following methods:

The frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy;

The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio;

The time-domain stability characteristic parameter is the expected ratio of the variance of the amplitude superposition value to the square of the amplitude superposition value, or the ratio is multiplied by a coefficient;

The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of certain spectral amplitudes, or the ratio is multiplied by a coefficient;

The tonality feature parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficient of the two frame signals before and after, or by continuing to smooth and filter the correlation value.

As shown in FIG. 5, Embodiment 2 of the activation sound detection (VAD) device of the present invention is different from Embodiment 1 in that the device also includes a sign calculation unit and a background noise energy processing unit, wherein:

The sign calculation unit is used to calculate the tonality sign of the current frame according to the frame energy parameter of the current frame, the characteristic parameter of the spectral center of gravity, the characteristic parameter of time domain stability, the characteristic parameter of spectral flatness, and the characteristic parameter of the tonality:

A background noise energy processing unit comprising:

The identification calculation module is used to calculate and obtain the initial background noise identification of the current frame according to the current frame frame energy parameters, spectral center of gravity characteristic parameters, time domain stability characteristic parameters, spectral flatness characteristic parameters, and tonality characteristic parameters;

The identification correction module is used to modify the initial background noise identification according to the decision result of the current frame VAD, the tonality characteristic parameter, the signal-to-noise ratio parameter, the tonality sign, and the time domain stability characteristic parameter;

The background noise energy acquisition module is used to obtain the background noise energy of the current frame according to the correction value of the background noise flag, the frame energy parameter of the current frame, and the full-band background noise energy of the previous frame, and the background noise energy of the current frame is used Calculate the SNR parameter in the next frame.

Corresponding to method embodiments 1 and 2, as shown in Figure 6, the VAD decision unit includes:

The long-term signal-to-noise ratio calculation module is used for the average long-term activation signal calculated by the previous frame

The ratio of the signal energy to the average long-term background noise energy is calculated to obtain the long-term signal-to-noise ratio lt_snr;

The average full-band signal-to-noise ratio calculation module is used to calculate the average value of the full-band signal-to-noise ratio SNR2 of several recent frames to obtain the average full-band signal-to-noise ratio SNR2_lt_ave;

The signal-to-noise ratio threshold calculation module is used to obtain the signal-to-noise ratio threshold snr_thr of the VAD decision according to the characteristic parameter of the spectral center of gravity, the long-term signal-to-noise ratio lt_snr, the number of continuous activation sound frames continuous_speech_num and the number of continuous noise frames in front of the continuous_noise_num;

The initial VAD decision module is used to calculate the initial VAD decision according to the signal-to-noise ratio parameters SNR1 and SNR2 calculated according to the decision threshold snr_thr of VAD and the current frame;

The VAD result correction module corrects the VAD judgment result according to the tonality mark, the average full-band signal-to-noise ratio SNR2_lt_ave, the center of gravity of the spectrum, and the long-term signal-to-noise ratio lt_snr;

The activation tone maintenance frame correction module is used to modify and obtain the activation tone according to the judgment results of the previous frames, the long-term SNR lt_snr, the average full-band SNR SNR2_lt_ave, the SNR of the current frame, and the VAD judgment result of the current frame Keep the number of frames;

The VAD judgment module is configured to add an active tone hold according to the decision result of the current frame and the number of active tone hold frames num_speech_hangover, to obtain the VAD decision result of the current frame.

More preferably, the VAD decision unit also includes: an energy calculation module, which is used to calculate the average long-term activation tone signal energy E _fg according to the VAD initial decision result; and update the average long-term background noise energy E _bg according to the VAD decision result , the updated value is used for the next frame VAD decision.

The present invention also provides an embodiment of a background noise detection method, as shown in FIG. 7, the method includes:

Step 701: Obtain the subband signal and spectrum amplitude of the current frame;

Step 702: Calculate the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectrum amplitude according to the values of the frame energy parameter, spectral center of gravity characteristic parameter, and temporal stability characteristic parameter calculated from the subband signal;

Preferably, the frame energy parameter is a weighted superposition value or a direct superposition value of signal energy of each subband.

The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or a value obtained by smoothing the ratio.

The time domain stability parameter is an expected ratio between the variance of the frame energy amplitude and the square of the amplitude superposition value, or the ratio is multiplied by a coefficient.

The spectral flatness parameter is the ratio of the geometric mean and the arithmetic mean of certain spectrum amplitudes, or the ratio multiplied by a coefficient.

Specifically, step 701 and step 702 may use the same method as above, which will not be repeated here.

Step 703: Perform background noise detection according to the spectral center of gravity characteristic parameters, time-domain stability characteristic parameters, spectral flatness characteristic parameters, tonality characteristic parameters, and current frame energy parameters, and determine whether the current frame is background noise.

Preferably, if any of the following conditions is judged to be true, then it is judged that the current frame is not a noise signal:

The time domain stability parameter lt_stable_rate0 is greater than a set threshold value;

The smoothing filter value of the characteristic parameter value of the center of gravity of the spectrum in the first interval is greater than a set threshold value, and the time domain stability characteristic parameter value is also greater than a certain set threshold value;

The tonality characteristic parameter or its smooth-filtered value is greater than a set threshold value, and the time domain stability characteristic parameter lt_stable_rate0 value is greater than its set threshold value;

The spectral flatness characteristic parameters of each sub-band or the values after smoothing and filtering are smaller than the respective corresponding set threshold values;

Or, it is determined that the value of the frame energy parameter E _t1 is greater than the set threshold value E_thr1.

Specifically, it is assumed that the current frame is background noise.

In this embodiment, a background noise flag background_flag is used to indicate whether the current frame is background noise, and it is agreed that if the current frame is judged to be background noise, the background noise flag background_flag is set to 1, otherwise the background noise flag background_flag is set to 0.

Whether the current frame is a noise signal is detected according to the time-domain stability characteristic parameter, the spectral center of gravity characteristic parameter, the spectral flatness characteristic parameter, the tonality characteristic parameter, and the current frame energy parameter. If it is not a noise signal, set the background noise flag background_flag to 0.

The specific process is as follows:

Determine whether the time domain stability parameter lt_stable_rate0 is greater than a set threshold lt_stable_rate_thr1. If yes, it is judged that the current frame is not a noise signal, and background_flag is set to 0. The value range of the threshold value lt_stable_rate_thr1 in this embodiment is [0.8, 1.6];

It is judged whether the characteristic parameter value of the center of gravity of the smooth spectrum is greater than a set threshold value sp_center_thr1, and the value of the characteristic parameter value of the time domain stability is also greater than a certain set threshold value lt_stable_rate_thr2. If yes, it is judged that the current frame is not a noise signal, and background_flag is set to 0. The value range of sp_center_thr1 is [1.6, 4]; the value range of 1t_stable_rate_thr2 is (0, 0.1].

Determine whether the value of the tonality characteristic parameter tonality_rate2 is greater than a set threshold value tonality_rate_thr1, whether the time domain stability characteristic parameter lt_stable_rate0 value is greater than the set threshold value lt_stable_rate_thr3, and if the above conditions are satisfied at the same time, it is judged that the current frame is not background noise , background_flag is assigned a value of 0. The value range of the threshold value tonality_rate_thr1 is [0.4, 0.66]. The value range of the threshold lt_stable_rate_thr3 is [0.06, 0.3].

Determine whether the value of the spectral flatness characteristic parameter sSMR[0] is less than the set threshold value sSMR_thr1, determine whether the value of the spectral flatness characteristic parameter sSMR[1] is less than the set threshold value sSMR_thr2, and determine the spectral flatness characteristic parameter Whether the value of sSMR[2] is less than the set sSMR_thr3. If the above conditions are satisfied at the same time, it is determined that the current frame is not background noise. background_flag is assigned a value of 0. The value range of the threshold values sSMR_thr1, sSMR_thr2, and sSMR_thr3 is [0.88, 0.98]. Determine whether the value of the flatness characteristic parameter sSMR[0] is less than the set threshold value sSMR_thr4, determine whether the value of the spectral flatness characteristic parameter sSMR[1] is less than the set threshold value sSMR_thr5, and determine the spectral flatness characteristic parameter sSMR Whether the value of [1] is less than the set threshold sSMR_thr6. If any of the above conditions is true, it is determined that the current frame is not background noise. background_flag is assigned a value of 0. The value range of sSMR_thr4, sSMR_thr5, and sSMR_thr6 is [0.80, 0.92]

It is judged whether the value of the frame energy parameter E _t1 is greater than the set threshold value E_thr1 , and if the above condition is satisfied, it is judged that the current frame is not background noise. background_flag is assigned a value of 0. E_thr1 takes the value according to the dynamic range of the frame energy parameter.

If the current frame is not detected as background noise, it means that the current frame is background noise.

Corresponding to the above method, the present invention also provides a background noise detection device, as shown in Figure 8, the device includes:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

The characteristic parameter calculation unit is used to obtain the values of the frame energy parameter, the spectral center of gravity characteristic parameter, and the time domain stability characteristic parameter calculated according to the subband signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude ;

Preferably, the frame energy parameter is a weighted superposition value or a direct superposition value of signal energy of each subband.

The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or a value obtained by smoothing the ratio.

The time domain stability parameter is an expected ratio between the variance of the frame energy amplitude and the square of the amplitude superposition value, or the ratio is multiplied by a coefficient.

The spectral flatness parameter is the ratio of the geometric mean and the arithmetic mean of certain spectrum amplitudes, or the ratio multiplied by a coefficient.

The background noise judging unit is used to detect the background noise according to the spectral center of gravity characteristic parameter, time domain stability characteristic parameter, spectral flatness characteristic parameter, tonality characteristic parameter, and current frame energy parameter, and judge whether the current frame is background noise.

Preferably, the background noise judging unit judges that any of the following conditions is true, then judges that the current frame is not a noise signal:

The time domain stability parameter lt_stable_rate0 is greater than a set threshold value;

The smoothing filter value of the characteristic parameter value of the center of gravity of the spectrum in the first interval is greater than a set threshold value, and the time domain stability characteristic parameter value is also greater than a certain set threshold value;

The tonality characteristic parameter or its smooth-filtered value is greater than a set threshold value, and the time domain stability characteristic parameter lt_stable_rate0 value is greater than its set threshold value;

The spectral flatness characteristic parameters of each sub-band or the values after smoothing and filtering are smaller than the respective corresponding set threshold values;

Or, it is determined that the value of the frame energy parameter E _t1 is greater than the set threshold value E_thr1.

The present invention also provides a method for detecting a tone signal, as shown in FIG. 9 , the method includes:

Step 901: Obtain the subband signal and spectrum amplitude of the current frame;

Step 902: Calculate and obtain the values of the spectral center of gravity characteristic parameter and time domain stability characteristic parameter of the current frame according to the sub-band signal, and obtain the values of the spectral flatness characteristic parameter and tonality characteristic parameter according to the spectrum amplitude;

Preferably, the characteristic parameter of the center of gravity of the spectrum is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing and filtering the ratio; the characteristic parameter of the time domain stability is the amplitude The expected ratio of the variance of the overlay to the square of the magnitude overlay, or multiply the ratio by a factor;

The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of certain spectral amplitudes, or the ratio is multiplied by a coefficient;

The tonality feature parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficient of the two frame signals before and after, or by continuing to smooth and filter the correlation value.

Step 903: Determine whether the current frame is a tonality signal according to the characteristic parameters of tonality, temporal stability, spectral flatness, and spectral center of gravity.

Step 903: When judging whether it is a tone signal, perform the following operations:

A) Assume that the current frame signal is an atonality signal, and use a tonality frame flag tonality_frame to indicate whether the current frame is a tonality frame.

In this embodiment, the value of tonality_frame is 1, indicating that the current frame is a tonal frame, and 0, indicating that the current frame is an atonality frame;

B) judging whether the value of tonality_ratel or tonality_rate2 after smoothing and filtering is greater than the corresponding set threshold value tonality_decision_thr1 or tonality_decision_thr2, if one of the above conditions is established, execute step C), otherwise execute step D);

Among them, the value range of tonality_decision_thr1 is [0.5, 0.7], and the value range of tonality_ratel is [0.7, 0.99].

C If the time domain stability characteristic parameter value lt_stable_rate0 is less than a set threshold value lt_stable_decision_thr1; the spectral center of gravity characteristic parameter value sp_center[1] is greater than a set threshold value spc_decision_thr1, and the spectral flatness characteristic parameters of each subband are equal to Less than their corresponding preset threshold values, specifically, the spectral flatness characteristic parameter sSMR[0] is less than a set threshold value sSMF_decision_thr1 or sSMR[1] is less than a set threshold value sSMF_decision_thr2 or sSMR[2 ] is less than a set threshold value sSMF_decision_thr3; then it is judged that the current frame is a tonal frame, and the value of the tonality_frame flag is set to 1; otherwise, it is judged as a non-tonality frame, and the value of the tonality frame flag is set to 0. And proceed to step D.

Among them, the value range of the threshold lt_stable_decision_thr1 is [0.01, 0.25], spc_decision_thr1 is [1.0, 1.8], sSMF_decision_thr1 is [0.6, 0.9], sSMF_decision_thr2[0.6, 0.9], sSMF_decision_thr3[0.7, 0.98].

D) Update the tonality degree characteristic parameter tonality_degree according to the tonality frame flag tonality_frame, wherein the initial value of the tonality degree parameter tonality_degree is set when the activation sound detection device starts to work, and the value range is [0, 1]. In different cases, the calculation method of tonality_degree characteristic parameter is different:

If the current tonality frame flag indicates that the current frame is a tonality frame, the tonality_degree characteristic parameter tonality_degree is updated using the following equation:

tonality_degree = tonality_degree _-1 td_scale_A+td_scale_B;

Among them, tonality_degree _-1 is the characteristic parameter of the tonality degree of the previous frame. Its initial value ranges from [0, 1]. td_scale_A is the attenuation coefficient, and its value range is [0, 1]; td_scale_B is the accumulation coefficient, and its value range is [0, 1].

E) Judging whether the current frame is a tonality signal according to the updated tonality degree characteristic parameter tonality_degree, and setting the value of the tonality flag tonality_flag.

Specifically, if the tonality_degree characteristic parameter tonality_degree is greater than a certain threshold value, it is judged that the current frame is a tonal signal; otherwise, it is judged that the current frame is a non-tonality signal.

Corresponding to the aforementioned tonality signal detection method, the present invention also provides a tonality signal detection device, as shown in Figure 10, the detection device includes:

A filter bank for obtaining subband signals of the current frame;

A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame;

The characteristic parameter calculation unit is used to calculate the values of the current spectral center of gravity characteristic parameter and the time domain stability characteristic parameter according to the sub-band signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude value calculation;

As mentioned above, the characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio;

The time-domain stability characteristic parameter is the expected ratio of the variance of the amplitude superposition value to the square of the amplitude superposition value, or the ratio is multiplied by a coefficient;

The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of certain spectral amplitudes, or the ratio is multiplied by a coefficient;

The tonality feature parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficient of the two frame signals before and after, or by continuing to smooth and filter the correlation value.

The tonality signal judging unit is used for judging whether the current frame is a tonality signal according to the tonality characteristic parameter, the temporal stability characteristic parameter, the spectral flatness characteristic parameter, and the spectral center of gravity characteristic parameter.

As shown in Figure 11, the tonality signal judging unit includes:

The tonality signal initialization module is used to set the current frame signal as an atonality signal, and uses a tonality frame flag tonality_frame to indicate whether the current frame is a tonality frame;

The tonality characteristic parameter judging module is used for judging whether the value of the tonality characteristic parameter tonality_rate1 or tonality_rate2 after smoothing and filtering is greater than the corresponding set threshold value;

The tonality signal judging module is used to determine that the tonality characteristic parameter judging module is yes, if the time domain stability characteristic parameter value is less than a set threshold value; the spectral center of gravity characteristic parameter value is greater than a set threshold value Limit value, and the spectral flatness characteristic parameters of each sub-band are less than the respective preset threshold values; judge that the current frame is a tonality frame; judge whether the current frame is a tonality frame according to the calculated tonality_degree characteristic parameter tonality_degree Tonality signal, and when the tonality characteristic parameter judging module judges as no, it is used to judge whether the current frame is a tonality signal according to the updated tonality degree characteristic parameter tonality_degree, and set the value of the tonality flag tonality_flag;

The tonality degree parameter update module is used to update the tonality degree characteristic parameter tonality_degree according to the tonality frame flag when the value of the tonality characteristic parameter tonality_rate1 or tonality_rate2 after smoothing filtering is less than the corresponding set threshold value, wherein The initial value of the tonality degree parameter tonality_degree is set when the activation sound detection device starts to work.

Specifically, if the current tonality frame flag indicates that the current frame is a tonality frame, the tonality degree parameter update module uses the following equation to update the tonality_degree characteristic parameter:

tonality_degree = tonality_degree _-1 td scale_A + td_scale_B;

Among them, tonality_degree _-1 is the characteristic parameter of the tonality degree of the previous frame. Its initial value ranges from [0, 1]. td_scale_A is the attenuation coefficient, and its value range is [0, 1]; td_scale_B is the accumulation coefficient, and its value range is [0, 1].

If the tonality degree characteristic parameter tonality_degree is greater than a certain threshold value, the tonality signal judging module judges that the current frame is a tonality signal, otherwise, judges that the current frame is an atonality signal.

Specifically, if the tonality_degree characteristic parameter tonality_degree is greater than the threshold value 0.5, then it is judged that the current frame is a tonal signal, and the value of the tonality flag tonality_flag is set to 1; otherwise, it is judged that the current frame is a non-tonality signal, and the value is set is 0. The range of threshold value for tone signal judgment is [0.3, 0.7].

The present invention also provides a method for correcting the frame number of the active tone in the VAD decision, as shown in Figure 12, the method includes:

Step 1201: Calculate the long-term signal-to-noise ratio lt_snr according to the sub-band signal;

Specifically, the long-term signal-to-noise ratio lt_snr is calculated through the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame; the long-term signal-to-noise ratio lt_snr can be expressed in logarithm.

Step 1202: Calculate the average full-band SNR SNR2_lt_ave;

Calculate the average value of the full-band signal-to-noise ratio SNR2 of several recent frames to obtain the average full-band signal-to-noise ratio SNR2_lt_ave;

Step 1203: According to the judgment results of the previous frames, the long-term SNR lt_snr, the average full-band SNR SNR2_lt_ave, the SNR parameters of the current frame, and the VAD judgment results of the current frame, the number of frames to maintain the current active tone is corrected .

Understandably, the precondition for maintaining the correction of the frame number of the current active sound is that the active sound flag indicates that the current frame is the active sound frame.

Preferably, when modifying the number of frames maintained for the current activation sound, if the previous continuous speech frame number is less than a set threshold value 1, and the long-term signal-to-noise ratio lt_snr is less than a set threshold value 2, the current active tone The number of sound holding frames is equal to the minimum number of continuous active sound frames minus the number of previous continuous speech frames; otherwise, if the average full-band SNR SNR2_lt_ave is greater than a set threshold value 3, and the number of previous continuous speech frames is greater than a set If the threshold value is 4, set the value of the number of frames to hold the active voice according to the long-term signal-to-noise ratio, otherwise the value of the number of frames to hold the current active voice num_speech_hangover will not be corrected.

Corresponding to the method for correcting the number of frames to be kept in the activation tone, the present invention also provides a correction device for maintaining the number of frames in the activation tone in the VAD decision. As shown in FIG. 13 , the correction device includes:

A long-term signal-to-noise ratio calculation unit, used to calculate the long-term signal-to-noise ratio lt_snr;

Specifically, the long-term signal-to-noise ratio calculation unit calculates the long-term signal-to-noise ratio lt_snr through the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame;

The average full-band SNR calculation unit is used to calculate the average full-band SNR SNR2_lt_ave;

Specifically, the average full-band signal-to-noise ratio calculation unit calculates the average value of the full-band signal-to-noise ratio SNR2 of several recent frames to obtain the average full-band signal-to-noise ratio SNR2_lt_ave.

Activation tone keeps the frame number correction unit, is used for according to the decision result of previous several frames, long-term signal-to-noise ratio lt_snr, average full-band signal-to-noise ratio SNR2_lt_ave, the signal-to-noise ratio parameter of current frame and the VAD decision result of current frame, to current Activation sound maintains frame number for correction.

As mentioned above, the precondition for maintaining the correction of the frame number of the current active sound is that the active sound flag indicates that the current frame is the active sound frame.

Preferably, when the activation tone maintenance frame number modification unit is to modify the current activation tone maintenance frame number, if the previous continuous speech frame number is less than a set threshold value 1, and the long-term signal-to-noise ratio lt_snr is less than a set Threshold value 2, the current active tone keeps the number of frames equal to the minimum number of continuous active tone frames minus the previous continuous speech frame number, otherwise, if the average full-band SNR SNR2_lt_ave is greater than a set threshold value 3, and the previous continuous If the number of speech frames is greater than a set threshold value of 4, then set the value of the number of frames to hold the activation tone according to the size of the long-term signal-to-noise ratio, otherwise the value of the number of frames of the current activation tone to hold nun_speech_hangover will not be corrected.

The present invention also provides a method for adjusting the SNR threshold in VAD decision, as shown in Figure 14, the adjustment method includes:

Step 1401: Calculate the characteristic parameter of the spectral center of gravity of the current frame according to the sub-band signal;

Specifically, the characteristic parameter of the spectral center of gravity is a ratio of a weighted cumulative value to an unweighted cumulative value of all or part of the sub-band signal energy, or a value obtained by smoothing the ratio.

Step 1402: Calculate the long-term signal-to-noise ratio lt_snr through the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame;

Step 1403: Adjust the SNR threshold of the VAD decision according to the characteristic parameters of the spectral center of gravity, the long-term SNR, the number of the previous continuous active audio frames and the number of the previous continuous noise frames continuous_noise_num.

Specifically, as shown in Figure 15, the steps of adjusting the SNR threshold include:

Step 1501: Set the initial value of the SNR threshold snr_thr;

Step 1502: adjust the value of the SNR threshold snr_thr for the first time according to the spectral center of gravity parameter;

Step 1503: adjust the value of the SNR threshold snr_thr twice according to the number of continuous_speech_num of the previous continuous activation sound frames, the number of continuous_noise_num of the previous continuous noise frames, the average full-band SNR SNR2_lt_ave and the long-term SNR lt_snr;

Step 1504: According to the value of the long-term SNR lt_snr, the SNR threshold snr_thr is finally corrected to obtain the SNR threshold snr_thr of the current frame.

Corresponding to the aforementioned adjustment method of the SNR threshold, the present invention also provides an adjustment device for the SNR threshold in VAD decision, as shown in FIG. 16 , the adjustment device includes:

A characteristic parameter acquisition unit, used to calculate the spectral center of gravity characteristic parameter of the current frame according to the subband signal;

Preferably, the characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or a value obtained by smoothing the ratio.

The long-term signal-to-noise ratio calculation unit is used to calculate the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated by the previous frame to calculate the long-term signal-to-noise ratio lt_snr;

The signal-to-noise ratio threshold adjustment unit is used to adjust the signal-to-noise ratio threshold of the VAD decision according to the characteristic parameters of the spectral center of gravity, the long-term signal-to-noise ratio, the number of continuous activation audio frames and the number of continuous_noise_num previous continuous noise frames.

Specifically, when the SNR threshold adjustment unit adjusts the SNR threshold, the initial value of the SNR threshold snr_thr is set; the value of the SNR threshold snr_thr is adjusted for the first time according to the spectral center of gravity parameter; Continuous_speech_num, the number of continuous noise frames before continuous_noise_num, the average full-band SNR SNR2_lt_ave and the long-term SNR lt_snr adjust the value of snr_thr twice; finally, according to the value of the long-term SNR lt_snr, adjust the SNR threshold snr_thr The final adjustment is to obtain the SNR threshold snr_thr of the current frame.

Many modern speech coding standards, such as AMR and AMR-WB, support the VAD function. In terms of efficiency, the VAD of these encoders does not perform well in all typical background noises. Especially under non-stationary noise, such as office noise, the VAD efficiency of these encoders is low. For music signals, these VADs sometimes have false detections, resulting in significant quality degradation of the corresponding processing algorithms.

The method of the invention overcomes the shortcomings of the existing VAD algorithm, and improves the accuracy of music detection while improving the detection efficiency of VAD for unstable noise. Therefore, the voice and audio signal processing algorithm using the VAD can obtain better performance.

The background noise detection method provided by the invention can make the estimation of the background noise more accurate and stable, and is beneficial to improving the accuracy of VAD detection. The tonality signal detection method provided by the invention improves the accuracy of tonality music detection. The method for correcting the frame number of the active tone maintained by the present invention can make the VAD algorithm better balanced in terms of performance and efficiency under different noises and signal-to-noise ratios. The method for adjusting the signal-to-noise ratio threshold in the VAD decision provided by the present invention can make the VAD decision algorithm achieve better accuracy under different signal-to-noise ratios, and further improve efficiency while ensuring quality.

Those skilled in the art can understand that all or part of the steps in the above method can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk or an optical disk, and the like. Optionally, all or part of the steps in the foregoing embodiments may also be implemented using one or more integrated circuits. Correspondingly, each module/unit in the foregoing embodiments may be implemented in the form of hardware, or may be implemented in the form of software function modules. The present invention is not limited to any specific combination of hardware and software.

Claims (20)

1. an activation tone detects VAD method, is characterized in that, the method comprises: Obtain the subband signal and spectrum amplitude of the current frame; The values of the frame energy parameter, the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter of the current frame are calculated according to the sub-band signal; the values of the spectral flatness characteristic parameter and the tonality characteristic parameter are obtained according to the spectral amplitude calculation; calculating the signal-to-noise ratio parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame and the signal-to-noise ratio sub-band energy; According to the frame energy parameters of the current frame, the characteristic parameters of the spectral center of gravity, the characteristic parameters of the temporal stability, the characteristic parameters of the spectral flatness, and the characteristic parameters of the tonality, the tonality flag of the current frame is calculated; Calculate the VAD judgment result according to the tonality mark, signal-to-noise ratio parameter, spectral center of gravity characteristic parameter, and frame energy parameter; Wherein, the frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy; The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio is multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 2. The method according to claim 1, characterized in that, before or after obtaining the VAD decision result, the method further comprises: According to the frame energy parameters of the current frame, the characteristic parameters of the spectral center of gravity, the characteristic parameters of the temporal stability, the characteristic parameters of the spectral flatness, and the characteristic parameters of the tonality, the initial background noise identification of the current frame is calculated; After the VAD judgment result is obtained, the method further includes: modifying the initial background noise identifier according to the judgment result of the current frame VAD, the tonality characteristic parameter, the signal-to-noise ratio parameter, the tonality flag, and the time domain stability characteristic parameter; According to the correction value of the background noise mark, the frame energy parameter of the current frame, and the full-band background noise energy of the previous frame, the sub-band background noise energy and the full-band background noise energy of the current frame are obtained; The background noise energy of the current frame is used for calculating the SNR parameter of the next frame. 3. The method of claim 1, wherein, According to the tonality flag, signal-to-noise ratio parameter, spectral center of gravity feature parameter, and frame energy parameter, the VAD judgment result is calculated, and the calculation steps are as follows: a, The long-term signal-to-noise ratio is calculated by the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame; b. Calculate the average value of the full-band SNR2 of several recent frames to obtain the average full-band SNR SNR2_lt_ave; c. Obtain the signal-to-noise ratio threshold snr_thr of the VAD decision according to the characteristic parameters of the spectral center of gravity, the long-term signal-to-noise ratio lt_snr, the number of continuous activation sound frames in the front continuous_speech_num and the number of continuous noise frames in front of continuous_noise_num; d, calculating an initial VAD decision according to the VAD decision threshold snr_thr and a signal-to-noise ratio parameter, wherein the signal-to-noise ratio parameter includes a sub-band average signal-to-noise ratio SNR1 and a full-band signal-to-noise ratio SNR2; e, modify the VAD decision result according to the tonality flag, the average full-band signal-to-noise ratio SNR2_lt_ave, the characteristic parameter of the spectral center of gravity and the long-term signal-to-noise ratio lt_snr; f, according to the judgment results of several previous frames, the long-term signal-to-noise ratio lt_snr, the average full-band signal-to-noise ratio SNR2_lt_ave, the signal-to-noise ratio parameter of the current frame and the VAD judgment result of the current frame, modify the number of frames to keep the activation tone; g. Add the active tone hold according to the decision result of the current frame and the number of active tone hold frames num_speech_hangover to obtain the VAD decision result of the current frame. 4. The method according to claim 3, characterized in that: after the step d, also include calculating the average long-time activation tone signal energy E _fg according to the VAD initial decision result; after the step g, also include calculating according to the VAD decision result The average long-term background noise energy E _bg , the calculated value is used for the VAD decision of the next frame. 5. An activation tone detects a VAD device, characterized in that the device comprises: A filter bank for obtaining subband signals of the current frame; A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame; The characteristic parameter acquisition unit is used to calculate the frame energy parameter, spectral center of gravity characteristic parameter and time domain stability characteristic parameter value of the current frame according to the subband signal; obtain the spectral flatness characteristic parameter and tonality characteristic parameter according to the spectral amplitude calculation value; A sign calculation unit, used to calculate the tonality sign of the current frame according to the current frame frame energy parameter, spectral center of gravity characteristic parameter, temporal stability characteristic parameter, spectral flatness characteristic parameter, and tonality characteristic parameter; A signal-to-noise ratio calculation unit, which is used to calculate the signal-to-noise ratio parameter of the current frame according to the background noise energy estimated in the previous frame, the frame energy parameter of the current frame, and the signal-to-noise ratio sub-band energy; The VAD judgment unit is used to calculate and obtain the VAD judgment result according to the tonality mark, the signal-to-noise ratio parameter, the characteristic parameter of the spectral center of gravity, and the frame energy parameter; Wherein, the frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy; The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio is multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 6. activation tone detection VAD device as claimed in claim 5 is characterized in that, The device also includes a background noise energy processing unit comprising: The identification calculation module is used to calculate and obtain the initial background noise identification of the current frame according to the current frame frame energy parameters, spectral center of gravity characteristic parameters, time domain stability characteristic parameters, spectral flatness characteristic parameters, and tonality characteristic parameters; The identification correction module is used to modify the initial background noise identification according to the decision result of the current frame VAD, the tonality characteristic parameter, the signal-to-noise ratio parameter, the tonality sign, and the time domain stability characteristic parameter; The background noise energy acquisition module is used to obtain the background noise energy of the current frame according to the correction value of the background noise flag, the frame energy parameter of the current frame, and the full-band background noise energy of the previous frame, and the background noise energy of the current frame is used Calculate the SNR parameter in the next frame. 7. activation sound detection VAD device as claimed in claim 5 is characterized in that, described VAD judgment unit comprises: The long-term signal-to-noise ratio calculation module is used to calculate the long-term signal-to-noise ratio lt_snr by calculating the ratio of the average long-term activation tone signal energy and the average long-term background noise energy calculated in the previous frame; The average full-band signal-to-noise ratio calculation module is used to calculate the average value of the full-band signal-to-noise ratio SNR2 of several recent frames to obtain the average full-band signal-to-noise ratio SNR2_lt_ave; The signal-to-noise ratio threshold calculation module is used to obtain the signal-to-noise ratio threshold snr_thr of the VAD decision according to the characteristic parameter of the spectral center of gravity, the long-term signal-to-noise ratio lt_snr, the number of continuous activation sound frames continuous_speech_num and the number of continuous noise frames in front of the continuous_noise_num; The initial VAD judgment module is used to calculate the initial VAD judgment according to the SNR parameters calculated by the VAD decision threshold snr_thr and the current frame, wherein the SNR parameters include the sub-band average SNR SNR1 and the full-band signal-to-noise ratio Noise ratio SNR2; The VAD result correction module corrects the VAD judgment result according to the tonality flag, the average full-band signal-to-noise ratio SNR2_lt_ave, the characteristic parameter of the spectral center of gravity and the long-term signal-to-noise ratio lt_snr; The activation tone maintenance frame correction module is used to modify and obtain the activation tone according to the judgment results of the previous frames, the long-term SNR lt_snr, the average full-band SNR SNR2_lt_ave, the SNR of the current frame, and the VAD judgment result of the current frame Keep the number of frames; The VAD judgment module is configured to add an active tone hold according to the decision result of the current frame and the number of active tone hold frames num_speech_hangover, to obtain the VAD decision result of the current frame. 8. activation tone detection VAD device as claimed in claim 7, is characterized in that: described VAD judgment unit also comprises: energy calculation module, is used for according to VAD initial decision result, calculates average long-time activation tone signal energy E _fg ; And calculate the average long-term background noise energy E _bg according to the VAD decision result, and the calculated value is used for the VAD decision of the next frame. 9. A background noise detection method, characterized in that the method comprises: Obtain the subband signal and spectrum amplitude of the current frame; The values of the frame energy parameter, the spectral center of gravity characteristic parameter, and the time domain stability characteristic parameter calculated according to the sub-band signal are obtained, and the values of the spectral flatness characteristic parameter and the tonality characteristic parameter are calculated according to the spectrum amplitude; Perform background noise detection based on spectral center of gravity characteristic parameters, temporal stability characteristic parameters, spectral flatness characteristic parameters, tonality characteristic parameters, and current frame energy parameters to determine whether the current frame is background noise; Wherein, the frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy; The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The spectral flatness parameter is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 10. method as claimed in claim 9, is characterized in that: judge that any of following conditions holds true, then judge that current frame is not noise signal: The time domain stability characteristic parameter lt_stable_rate0 is greater than a set threshold value; The smoothing filter value of the characteristic parameter value of the center of gravity of the spectrum in the first interval is greater than a set threshold value, and the time domain stability characteristic parameter value is also greater than a certain set threshold value; The tonality characteristic parameter or its smooth-filtered value is greater than a set threshold value, and the time domain stability characteristic parameter lt_stable_rate0 value is greater than its set threshold value; The spectral flatness characteristic parameters of each sub-band or the values after smoothing and filtering are smaller than the respective corresponding set threshold values; Or, it is determined that the value of the frame energy parameter E _t1 is greater than the set threshold value E_thr1 . 11. A background noise detection device, characterized in that the device comprises: A filter bank for obtaining subband signals of the current frame; A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame; The characteristic parameter calculation unit is used to obtain the values of the frame energy parameter, the spectral center of gravity characteristic parameter, and the time domain stability characteristic parameter calculated according to the subband signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude ; The background noise judging unit is used to detect the background noise according to the spectral center of gravity characteristic parameter, time domain stability characteristic parameter, spectral flatness characteristic parameter, tonality characteristic parameter and current frame energy parameter, and judge whether the current frame is background noise; Wherein, the frame energy parameter is a weighted superposition value or a direct superposition value of each subband signal energy; The characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The spectral flatness parameter is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 12. The detection device according to claim 11, wherein the background noise judging unit judges that any of the following conditions is established, and then judges that the current frame is not a noise signal: The time domain stability characteristic parameter lt_stable_rate0 is greater than a set threshold value; The smoothing filter value of the characteristic parameter value of the center of gravity of the spectrum in the first interval is greater than a set threshold value, and the time domain stability characteristic parameter value is also greater than a certain set threshold value; The tonality characteristic parameter or its smooth-filtered value is greater than a set threshold value, and the time domain stability characteristic parameter lt_stable_rate0 value is greater than its set threshold value; The spectral flatness characteristic parameters of each sub-band or the values after smoothing and filtering are smaller than the respective corresponding set threshold values; Or, it is determined that the value of the frame energy parameter E _t1 is greater than the set threshold value E_thr1. 13. A tone signal detection method, characterized in that the method comprises: Obtain the subband signal and spectrum amplitude of the current frame; Calculate the values of the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter according to the sub-band signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude; Judging whether the current frame is a tonality signal according to the tonality characteristic parameter, the temporal stability characteristic parameter, the spectral flatness characteristic parameter, and the spectral center of gravity characteristic parameter; Wherein, the characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio is multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 14. The method according to claim 13, characterized in that: when judging whether it is a tone signal, perform the following operations: A) Assume that the current frame signal is an atonality signal, and use a tonality frame flag tonality_frame to indicate whether the current frame is a tonality frame; B) judging whether the value of the tonality characteristic parameter tonality_rate1 or tonality_rate2 after smooth filtering is greater than the corresponding set threshold value, if one of the above conditions is established, execute step C), otherwise execute step D); C) If the time-domain stability characteristic parameter value is less than a set threshold value; the spectral center of gravity characteristic parameter value is greater than a set threshold value, and the spectral flatness characteristic parameters of each subband are less than the corresponding preset threshold value; then judge that the current frame is a tonal frame, set the value of the tonal frame sign, otherwise it is judged as a non-tonal frame, set the value of the tonal frame sign, and continue to perform step D); D) update the tonality degree characteristic parameter tonality_degree according to the tonality frame mark, wherein the initial value of the tonality degree parameter tonality_degree is set when the activation sound detection starts to work; E) Judging whether the current frame is a tonality signal according to the updated tonality degree characteristic parameter tonality_degree, and setting the value of the tonality flag tonality_flag. 15. The method according to claim 14, characterized in that: if the current tonality frame sign indicates that the current frame is a tonality frame, then the tonality_degree characteristic parameter tonality_degree is updated using the following equation: tonality_degree = tonality_degree _-1 td_scale_A+td_scale_B; Among them, tonality_degree _-1 is the tonality degree characteristic parameter of the previous frame, and its initial value ranges from [0,1], td_scale_A is the attenuation coefficient, and td_scale_B is the accumulation coefficient. 16. The method of claim 14, wherein: If the tonality degree characteristic parameter tonality_degree is greater than a certain threshold value, it is judged that the current frame is a tonality signal, otherwise, it is judged that the current frame is a non-tonality signal. 17. A tone signal detection device, characterized in that the detection device comprises: A filter bank for obtaining subband signals of the current frame; A spectrum amplitude calculation unit, configured to obtain the spectrum amplitude of the current frame; The characteristic parameter calculation unit is used to calculate the values of the spectral center of gravity characteristic parameter and the time domain stability characteristic parameter according to the sub-band signal, and obtain the values of the spectral flatness characteristic parameter and the tonality characteristic parameter according to the spectral amplitude calculation; A tonality signal judging unit, configured to judge whether the current frame is a tonality signal according to the tonality characteristic parameter, the time domain stability characteristic parameter, the spectral flatness characteristic parameter, and the spectral center of gravity characteristic parameter; Wherein, the characteristic parameter of the spectral center of gravity is the ratio of the weighted cumulative value and the unweighted cumulative value of all or part of the sub-band signal energy, or the value obtained by smoothing the ratio; The time-domain stability characteristic parameter is the expected ratio between the variance of the energy amplitude superposition value and the square of the energy amplitude superposition value, or the ratio multiplied by a coefficient; The characteristic parameter of spectral flatness is the ratio of the geometric mean and the arithmetic mean of the spectral amplitudes of multiple subbands, or the ratio is multiplied by a coefficient; The tonality characteristic parameter is obtained by calculating the correlation value of the intra-frame spectral difference coefficients of two frames of signals before and after, or by continuing to smooth and filter the correlation value. 18. The detection device according to claim 17, characterized in that: the tonality signal judging unit comprises: The tonality signal initialization module is used to set the current frame signal as an atonality signal, and uses a tonality frame flag tonality_frame to indicate whether the current frame is a tonality frame; The tonality characteristic parameter judging module is used for judging whether the value of the tonality characteristic parameter tonality_rate1 or tonality_rate2 after smoothing and filtering is greater than the corresponding set threshold value; The tonality signal judging module is used to determine that the tonality characteristic parameter judging module is yes, if the time domain stability characteristic parameter value is less than a set threshold value; the spectral center of gravity characteristic parameter value is greater than a set threshold value Limit value, and the spectral flatness characteristic parameters of each sub-band are less than the respective preset threshold values; judge that the current frame is a tonality frame; judge whether the current frame is a tonality frame according to the calculated tonality_degree characteristic parameter tonality_degree Tonality signal, and when the tonality characteristic parameter judging module judges as no, it is used to judge whether the current frame is a tonality signal according to the updated tonality degree characteristic parameter tonality_degree, and set the value of the tonality flag tonality_flag; The tonality degree parameter update module is used to update the tonality degree characteristic parameter tonality_degree according to the tonality frame flag when the value of the tonality characteristic parameter tonality_rate1 or tonality_rate2 after smoothing filtering is less than the corresponding set threshold value, wherein The initial value of the tonality degree parameter tonality_degree is set when the activation sound detection device starts to work. 19. The detection device according to claim 18, wherein if the current tonality frame flag indicates that the current frame is a tonality frame, then the tonality degree parameter update module uses the following equation to update the tonality_degree characteristic parameter: tonality_degree = tonality_degree _-1 td_scale_A+td_scale_B; Among them, tonality_degree _-1 is the tonality degree characteristic parameter of the previous frame, and its initial value ranges from [0,1], td_scale_A is the attenuation coefficient, and td_scale_B is the accumulation coefficient. 20. The detection device as claimed in claim 18, characterized in that: if the tonality_degree characteristic parameter tonality_degree is greater than a certain set threshold value, then the tonality signal judging module judges that the current frame is a tonality signal, otherwise , judging that the current frame is an atonal signal.