CN101393744A - Method for regulating threshold and detection module - Google Patents

Abstract

The embodiment of the invention discloses a method for adjusting the threshold value, which comprises the following steps: determining the signal characteristics of the received signals; further determining whether continuous useless signals are generated, if that the useless signal characteristics are represented by the signals is determined, or further determining whether continuous useful signals are generated, if that the useful signal characteristics are represented by the signals is determined; reducing the threshold additional correction value if the continuous useful signals are generated; and taking the sum of the original threshold value used for comparing and the increased or reduced threshold additional correction value as the threshold value for comparing. Correspondingly, the embodiment of the invention further provides a detection module. The technical proposal of the embodiment of the invention can increase the correct rate of the signal decision.

Description

Adjust the method and the detection module of threshold value

Technical field

The present invention relates to the voice communication technical field, be specifically related to a kind of method and detection module of adjusting threshold value.

Background technology

In voice communication, the processing of voice is mainly finished by audio coder ﹠ decoder (codec).Because people's sounding is not continuous in voice communication, nearly about 70% free time is speech not, and therefore carrying out encoding and decoding speech with same speed all the time is a kind of great waste to communication channel resources.In view of this consideration, people have introduced voice activation detection (Voice Activity Detector, VAD) technology in speech coder.This technology can effectively have been distinguished language sound (audible signal) and no language sound (no acoustical signal), when being arranged, the language sound encodes with normal code rate, form speech frame, and there is not the language sound is some ground unrests and quiet, therefore can handle with simple coded system, form quiet description (Silence Descriptor, SID) frame.By this processing mode, can make scrambler when guaranteeing signal encoding quality (audible signal is encoded with higher code rate), obviously reduced the transmission bandwidth (no acoustical signal is encoded with lower code rate) of signal again.

In addition, because the signal in the communications has been not only voice, also comprise various useful sound signals, as music, song, signal tone etc., so for convenience, the back is referred to as useful signal with the audible signal in these and the voice, useless signals such as ground unrest are referred to as garbage signal, and useful signal and garbage signal is commonly referred to as signal of communication.And the function of VAD module also has been not only to be used for detecting voice signal, but be used for detecting useful signal, also promptly become voice activation detection (SAD) so voice activation detects (VAD), in order to keep consistency with primal algorithm when introducing technical scheme, this paper still uses VAD when description.

Third generation partner program 3GPP (the 3rd Generation Partnership Project) organized to set up a series of at but be not limited to the multi-rate coding standard of third generation communication system, for example self-adaption multi-speed encoding/decoding algorithm AMR (Adaptive Multi-Rate), AMR-WB code decode algorithm AMR-WB (Adaptive Multi-Rate-Wideband), AMR-WB+ coding standard, they also all contain the VAD module.The judgement main logic of its VAD module all is the whole signal to noise ratio (S/N ratio) of current demand signal and an adaptive decision threshold to be compared decide whether current demand signal is useful signal: if signal to noise ratio (S/N ratio) is bigger than decision threshold, think that so current demand signal is a useful signal; Otherwise just think that current demand signal is a garbage signal.Therefore decision threshold can produce directly influence to court verdict, thereby has influence on coding quality and code rate.

Speech coder AMR has two kinds of VAD modules: VAD1 module and VAD2 module, both differences are not very big, below are introduced with the testing process of VAD2 module to signal.

Seeing also Fig. 1, is that the VAD2 module comprises the testing process figure of signal among the prior art one speech coder AMR:

Step 101, judge signal to noise ratio (S/N ratio) v (m) whether greater than decision threshold and its modified value and vth+negSNRbias, if not, enter step 102, if enter step 104;

The signal to noise ratio (S/N ratio) of each subband to the speech frame of input quantizes and sues for peace to obtain the signal to noise ratio (S/N ratio) of this speech frame $v\left(m\right)=\underset{i=0}{\overset{N_c-1}{\mathrm{\Σ}}}V\left({\mathrm{\σ}}_q\left(i\right)\right),$ Wherein V () is a quantizer, and Nc=16 is the number of subband, and 0≤i≤Nc, m are frame numbers.

Y-PSNR SNR during according to signal of communication long _qTable look-up and obtain VAD decision threshold v _Th: v _Th=v _Table(SNR _q), work as SNR _qThan hour v _ThAlso less, work as SNR _qV when big _ThAlso bigger.

Calculate decision threshold modified value negSNRbias (m):

$\mathrm{negSNRbias}\left(m\right)=\left\{\begin{array}{c}\max \{12\&times;\mathrm{negSNRvar}\left(m\right)-\mathrm{0.65,0}\},\mathrm{SNR}<0\\ \mathrm{negSNRbias}(m-1),\mathrm{otherwise}\end{array}\right.$

Wherein $\mathrm{negSNRvar}\left(m\right)=\left\{\begin{array}{c}\min \{0.99\mathrm{negSNRvar}(m-1)+0.01{\mathrm{SNR}}^2\mathrm{,4.0}\},\mathrm{SNR}<0\\ \mathrm{negSNRvar}(m-1),\mathrm{otherwise}\end{array}\right.,$

NegSNRbias (m) is by the decision of noise stability state, and when astable noise occurring, it is very low that SNR can become, and this moment, thresholding modified value negSNRbias can increase.

After obtaining signal to noise ratio (S/N ratio) v (m), decision threshold vth and decision threshold modified value negSNRbias, judge that whether v (m) is greater than vth+negSNRbias, if not, VAD=0 judges that tentatively signal is a garbage signal, enters step 102, if, VAD=1, the judgement signal is a useful signal, enters step 104.

Step 102 judges whether to be in the hangover stage, if not, enters step 103, if enter step 104;

If judging signal is not in the hangover stage, enter step 103, if judging signal is in the hangover stage, enters step 104 VAD=0 is modified to VAD=1.

Because some low-energy useful signals were difficult to directly be detected when useful signal finished, therefore having carried out hangover after initial VAD judgement handles, promptly after initial VAD judgement testing result, the signal when if signal is end, the court verdict that with initial court verdict is several frame signals of 0 is modified to 1, some low-yield useful signals when so just effectively detecting the useful signal end can effectively guarantee the coding quality that accurately reaches useful signal that VAD adjudicates.

Step 103, be not in the hangover stage, therefore finally judge VAD=0, signal is a garbage signal;

Step 104, judge VAD=1, signal is a useful signal.

The VAD technology of AMR-WB also is a kind of comparatively ripe and reliable VAD technology.

Seeing also Fig. 2, is that the VAD module comprises the testing process figure of signal among prior art two AMR-WB:

Step 201, judge that signal to noise ratio (S/N ratio) snr_sum whether greater than decision threshold vad_thr, if not, enters step 202, if enter step 204;

The signal to noise ratio (S/N ratio) snr_sum of current speech frame is:

$\mathrm{snr}\_\mathrm{sum}=\underset{n=1}{\overset{12}{\mathrm{\&Sigma;}}}\mathrm{MAX}{(1.0,\frac{\mathrm{level}\left[n\right]}{\mathrm{bckr}\_\mathrm{est}\left[n\right]})}^2,$ Bckr_est[n wherein] be the ground unrest estimated value of each subband.

Decision threshold vad_thr=NO_SLOPE* (ilog2_noise_level-NO_P1)+THR_HIGH+MIN (SP_CH_MAX, MAX (SP_CH_MIN, SP_CH_MIN+SP_SLOPE* (ilog2_speech_level-SP_P1)))

NO_SLOPE wherein, SP_SLOPE, NO_P1, SP_P1, THR_HIGH, SP_CH_MAX and SP_CH_MIN all are constants, ilog2_noise_level is that the estimated value of compute level useful signal and taking the logarithm obtains.Therefore, decision threshold vad_thr adjusts according to the estimated value of useful signal level and the estimated value self-adaptation of background-noise level.

After obtaining signal to noise ratio (S/N ratio) snr_sum and decision threshold vad_thr, whether judge signal to noise ratio (S/N ratio) snr_sum greater than decision threshold vad_thr, if not, VAD=0, preliminary judgement signal is a garbage signal, enters step 202, if, VAD=1, the judgement signal is a useful signal, enters step 204.

Step 202 judges whether to be in the hangover stage, if not, enters step 203, if enter step 204;

If judging signal is not in the hangover stage, enter step 203, if judging signal is in the hangover stage, enters step 204 VAD=0 is modified to VAD=1.

Because some low-energy useful signals were difficult to directly be detected when useful signal finished, therefore having carried out hangover after initial VAD judgement handles, promptly after initial VAD judgement testing result, the signal when if signal is end, the court verdict that with initial court verdict is several frame signals of 0 is modified to 1, some low-yield useful signals when so just effectively detecting the useful signal end can effectively guarantee the coding quality that accurately reaches useful signal that VAD adjudicates.

Step 203, be not in the hangover stage, therefore finally judge VAD=0, signal is a garbage signal;

Step 204, judge VAD=1, signal is a useful signal.

In research and practice process to prior art, the inventor finds that there is following problem in prior art:

In the prior art one, under the situation of noise that continues or slight astable noise, decision threshold modified value negSNRbias inoperative substantially (being worth very little), will make that like this vth+negSNRbias that compares with signal to noise ratio (S/N ratio) under the ground unrest situation is on the low side, make the result of VAD judgement a lot " burr " (noise have been judged to into useful signal) occur, thereby cause the waste of bandwidth;

In the prior art two, ending in voice or music, because Signal-to-Noise snr_sum is less relatively, and decision threshold vad_thr does not have significant change, therefore erroneous judgement will appear in the VAD judgement, can judge into these useful signal endings by accident garbage signal, also promptly can under the situation that lasting useful signal occurs, judge into the part useful signal by accident garbage signal, thereby cause the decline of coding quality because decision threshold is too high.

Summary of the invention

The technical matters that the embodiment of the invention will solve provides a kind of method and detection module of adjusting threshold value, can improve the accuracy of signal decision.

For solving the problems of the technologies described above, embodiment provided by the present invention is achieved through the following technical solutions:

The embodiment of the invention provides a kind of method of adjusting threshold value, comprising: the characteristics of signals of judging the signal that receives; If judging described characterization goes out the garbage signal characteristic, further judge whether to occur continuous garbage signal, if judging described characterization goes out the useful signal characteristic, further judge whether to occur continuous useful signal; If continuous garbage signal then increases thresholding additional corrections value, if continuous useful signal then reduces thresholding additional corrections value; The former threshold value that is useful on comparison added the above increase or reduce after thresholding additional corrections value as the threshold value that compares.

The embodiment of the invention provides a kind of detection module, comprising: judging unit is used to judge the characteristics of signals of the signal of reception; First processing unit, be used for when described judgment unit judges goes out described characterization and goes out the garbage signal characteristic, further judge whether to occur continuous garbage signal, when described judgment unit judges goes out described characterization and goes out the useful signal characteristic, further judge whether to occur continuous useful signal; Second processing unit is used for increasing thresholding additional corrections value when continuous garbage signal appears in described first processing unit judgement, when continuous useful signal appears in described first processing unit judgement, reduces thresholding additional corrections value; The 3rd processing unit is used to determine the threshold value that compares, with the former threshold value that is useful on comparison add increase in the above second processing unit or reduce after thresholding additional corrections value as the threshold value that compares.

Above technical scheme as can be seen, because embodiment of the invention technical scheme is the situation that continues appearance according to different characteristic between useful signal and the garbage signal, judge whether to occur continuous useful signal or garbage signal, thereby self-adaptation increases or reduces thresholding additional corrections value, on the basis of original judgment thresholding suitable adjustment whole decision threshold, thereby solved in the prior art a large amount of signal misjudged problems under ground unrest situation that continues and lasting useful signal situation, improved the accuracy of signal decision.

Description of drawings

Fig. 1 be among the prior art one speech coder AMR VAD2 module to the testing process figure of signal;

Fig. 2 be among prior art two AMR-WB VAD module to the testing process figure of signal;

Fig. 3 is the process flow diagram of embodiment of the invention correction decision threshold additional corrections value;

The process flow diagram of Fig. 4 to be the embodiment of the invention one revise when the VAD2 module detects signal among AMR decision threshold additional corrections value;

Fig. 5 revises the process flow diagram of decision threshold additional corrections value when to be the embodiment of the invention two at the VAD module of AMR-WB detect signal;

Fig. 6 is an embodiment of the invention detection module structural representation.

Embodiment

The embodiment of the invention provides a kind of method of adjusting threshold value, can improve the accuracy of signal decision.

Embodiment of the invention method is provided with decision threshold additional corrections value, and the obvious difference on some features according to useful signal and garbage signal, decision threshold additional corrections value is controlled automatically and revised, decision threshold additional corrections value is compared with the signal to noise ratio (S/N ratio) of signal after in conjunction with the former threshold value that is useful on comparison, thereby reach the purpose that improves the signal decision accuracy.

Seeing also Fig. 3, is the process flow diagram of embodiment of the invention correction decision threshold additional corrections value, comprising:

Step 301, from the signal that receives, obtain the parameter of characterization signal characteristic;

Extract the characteristic parameter that can characterize useful signal and garbage signal characteristic according to input signal.Characteristic parameter can be that entropy, tone or other can be represented the parameter of useful signal and garbage signal feature.The characteristic parameter that extracts can be calculated by scrambler, and for example tone is exactly that scrambler produces, also can newly calculate, for example the entropy of signal calculated.

Step 302, what judge characterization is useful signal characteristic or garbage signal characteristic, if what characterize is the useful signal characteristic, enters step 303, if what characterize is the garbage signal characteristic, enters step 306;

Judge that according to characteristic parameter current is useful signal characteristic or garbage signal characteristic.If characteristic parameter is an entropy, entropy and a preset value need be compared, if greater than preset value, then characterize out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.If characteristic parameter is a tone, then the pitch marks that produces according to scrambler is judged, if the pitch marks of described generation is the sign that expression does not contain the tone of signal, then characterizes out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.

Step 303, having with event counter continuously of will being provided with are added up, and occur the garbage signal counter O reset continuously with what be provided with;

Be provided with in the embodiment of the invention and have with event counter continuously and occur these two counters of garbage signal counter continuously, if what step 302 was judged the characteristic parameter sign is the useful signal characteristic, having with event counter continuously of then will being provided with adds up, and occurs the garbage signal counter O reset continuously with what be provided with.

Step 304, judge whether to occur continuous useful signal, if, enter step 305, if not, enter step 309;

Judge according to described having continuously whether signal is in the state of continuous useful signal with event counter, if described having continuously with the value of event counter set number greater than second, then judge the state that is in continuous useful signal, enter step 305, otherwise enter step 309.

Step 305, decision threshold additional corrections value is reduced;

After decision threshold additional corrections value is reduced, again the former threshold value that is useful on comparison is added the above decision threshold additional corrections value that reduces as the threshold value that compares, thereby reduce whole decision threshold, to guarantee that useful signal is not judged to garbage signal.

Step 306, having continuously of will being provided with are used the event counter zero clearing, and the garbage signal counter that occurs continuously that is provided with is added up;

Be provided with in the embodiment of the invention and have with event counter continuously and occur these two counters of garbage signal counter continuously, if what step 302 was judged the characteristic parameter sign is the garbage signal characteristic, the event counter zero clearing is used in having continuously of then will being provided with, and the garbage signal counter that occurs continuously that is provided with is added up.

Step 307, judge whether to occur continuous garbage signal, if, enter step 308, if not, enter step 309;

Judge according to the described garbage signal counter that occurs continuously whether signal is in the state of continuous garbage signal, if the described value that occurs the garbage signal counter is continuously set number greater than first, then judge the state that is in continuous garbage signal, enter step 308, otherwise enter step 309.

Step 308, decision threshold additional corrections value is increased;

After decision threshold additional corrections value is increased, again the former threshold value that is useful on comparison is added the decision threshold additional corrections value of the above increase as the threshold value that compares, thereby improve whole decision threshold, to guarantee that garbage signal is not judged to useful signal.

Step 309, maintenance decision threshold additional corrections are worth constant.

If do not have continuous useful signal characteristic or garbage signal characteristic, then do not revise decision threshold additional corrections value.Afterwards, the more former threshold value that is useful on comparison is added the decision threshold additional corrections value of the above increase as the threshold value that compares.

Below introduce the specific embodiment of the invention in detail.

See also Fig. 4, the process flow diagram of decision threshold additional corrections value that to be the embodiment of the invention one revise when the VAD2 module detects signal among the AMR.

Adopt this characteristic parameter of entropy of signal to distinguish useful signal and garbage signal among this embodiment one.Garbage signal refers generally to noise signal, so present embodiment adopts noise signal to be described.Entropy has been represented the uncertain degree of signal, and the uncertain degree of signal is high more, and entropy is big more.Generally speaking, the uncertain degree of noise is greater than the uncertain degree of useful signal, so if entropy during greater than the certain predetermined value, can think that signal has the obvious noise characteristic.This embodiment when being continuous ground unrest characteristic, increasing decision threshold additional corrections value, thereby improves whole decision threshold to current according to the size detection of entropy, reduces the False Rate of noise, improves whole judgement accuracy rate.

This embodiment sets in advance a decision threshold additional corrections value HbSNRbias based on entropy, and initial value generally gets 0.

Specifically comprise step among Fig. 4:

After step 401, the received signal, calculate the entropy hb of current demand signal;

$\mathrm{hb}\left(m\right)=-\underset{i}{\mathrm{\&Sigma;}}p(m,i)\lg \left(p(m,i)\right)$

Wherein $p(m,i)=E_{\mathrm{ch}}(m,i)/\underset{i}{\mathrm{\&Sigma;}}{E}_{\mathrm{ch}}(m,i),$ E _Ch(m i) is the normalized energy of subband i, and 0≤i≤Nc, Nc=16 are the numbers of subband, and m is a frame number.

Step 402, the entropy hb of signal is done level and smooth when long, obtain the entropy mean_hb after level and smooth; Mean_hb (m)=0.95mean_hb (m-1)+0.05hb (m)

Step 403, whether judge entropy mean_hb after level and smooth greater than preset value thr, if not, sign be the useful signal characteristic, enter step 404, if, sign be the noise signal characteristic, enter step 407;

A mean_hb and a preset value thr are compared, observe the noisiness situation in the certain phase.Preset value is value rule of thumb, gets thr=3.2 in the present embodiment.If mean_hb, thinks then that the obvious noise characteristic is arranged greater than thr, otherwise think the useful signal characteristic.

Step 404, will hang down entropy counter Hb_low_continue_counter continuously and add up 1, simultaneously with continuous high entropy counter Hb_high_continue_counter zero clearing;

Be provided with continuous high entropy counter Hb_high_continue_counter and continuously low entropy counter Hb_low_continue_counter in the embodiment of the invention, if what step 403 was judged the characteristic parameter sign is the useful signal characteristic, then will hang down entropy counter Hb_low_continue_counter continuously and add up 1, simultaneously with continuous high entropy counter Hb_high_continue_counter zero clearing.

Step 405, judge continuously low entropy counter Hb_low_continue_counter count value whether greater than number M, if, enter step 406, if not, enter step 410;

If the count value of Hb_low_continue_counter is greater than M, promptly the value of the mean_hb of continuous N frame is thought the useful signal characteristic so all less than thr, enters step 406, otherwise enters step 410.Number M is value rule of thumb, gets M=5 among this embodiment.

Step 406, with the value of the HbSNRbias k that successively decreases, k is value rule of thumb, gets k=2 among this embodiment;

Step 407, will hang down entropy counter Hb_low_continue_counter zero clearing continuously, simultaneously continuous high entropy counter Hb_high_continue_counter be added up 1;

Step 408, judgement are high continuously

Whether the count value of value counter t48Hb_high_continue_counter greater than number N, if, enter step 409, if not, enter step 410;

If the count value of Hb_high_continue_counter is greater than N, promptly the value of the mean_hb of N continuous frame is thought the characteristic of noise so all greater than thr, enters step 409, otherwise enters step 410.Number N is value rule of thumb, gets N=10 among this embodiment.

Step 409, the value of HbSNRbias is increased progressively h, h is value rule of thumb, gets h=1 among this embodiment;

Step 409, keep value constant of HbSNRbias.

Draw the value of HbSNRbias according to above-mentioned steps after, the value of HbSNRbias is added in does comprehensive judgement on the VAD decision threshold, promptly the conclusive judgement thresholding is vth+negSNRbias+HbSNRbias, and then operates by existing follow-up flow process.

Need to prove, can the end value of HbSNRbias be limited, make it in rational mobility scale, this mobility scale can certain negative value and on the occasion of between, to avoid the excessive or too small situation of this value appearance.

It should be noted that also above-mentioned said preset value thr, M, N, h etc. are the parameter of value rule of thumb, and value can be set as the case may be flexibly.

Can find, the characteristic that the entropy after level and smooth is observed input signal when long in the present embodiment, if it is all very big to have detected continuous entropy, think that then noisiness is obvious, just thresholding additional corrections value is increased with slower speed, thereby improve whole decision threshold, guarantee that noise is not mistaken for useful signal; When having detected continuous entropy when all very little, think that then the useful signal characteristic is obvious, thresholding additional corrections value is reduced with fast speeds, be not mistaken for noise to guarantee useful signal.Therefore according to the feature of input signal, whole decision threshold is carried out up or down adjustment, improved the accuracy of VAD judgement.

See also Fig. 5, revise the process flow diagram of decision threshold additional corrections value when to be the embodiment of the invention two at the VAD module of AMR-WB detect signal.

Adopt this characteristic parameter of tone of signal to distinguish useful signal and garbage signal among this embodiment.Garbage signal refers generally to noise signal, so present embodiment adopts noise signal to be described.Comprised tangible tone characteristic in the useful signal, and had continuous tone information, noise signal does not then have continuous tone characteristic.The continuity that this embodiment occurs according to tone information, judging current is useful signal feature or noise signal feature, and thresholding additional corrections value is controlled, and adjusts whole decision threshold, improves the judgement accuracy.

This embodiment sets in advance a decision threshold additional corrections value toneSNRbias based on tone, and initial value generally gets 0.

Specifically comprise step among Fig. 5:

Step 501, judged whether tone, if, enter step 502, if not, enter step 505;

The pitch marks tone_flag that produces according to the AMR-WB scrambler judges whether current frame signal has tone information, if pitch marks tone_flag represents to contain tone with 1, represent not contain tone with 0, then pitch marks tone_flag entered step 502 at 1 o'clock, and tone_flag entered step 505 at 0 o'clock.

Step 502, continuous tone counter tone_continue_counter is added up 1, will not have tone counter nonetone_continue_counter zero clearing simultaneously continuously;

Be provided with continuous no tone counter nonetone_continue_counter and continuous tone counter tone_continue_counter in the embodiment of the invention, if what step 502 was judged the characteristic parameter sign is the useful signal characteristic, then continuous tone counter tone_continue_counter is added up 1, will not have tone counter nonetone_continue_counter zero clearing simultaneously continuously.

Step 503, judge continuous tone counter tone_continue_counter count value whether greater than number N, if, enter step 504, if not, enter step 508;

If the count value of tone_continue_counter is greater than N, promptly the N continuous frame all has tone information, thinks the characteristic of useful signal so, enters step 504, otherwise enters step 508.Number N is value rule of thumb, gets N=5 among this embodiment.

Step 504, with the value of the toneSNRbias k that successively decreases, k is value rule of thumb, gets k=1 among this embodiment;

Step 505, with continuous tone counter tone_continue_counter zero clearing, will not have tone counter nonetone_continue_counter simultaneously continuously and add up 1;

Step 506, judge continuously no tone counter nonetone_continue_counter count value whether greater than number M, if, enter step 507, if not, enter step 508;

If the count value of nonetone_continue_counter is greater than M, promptly the continuous N frame does not all have tone information, thinks the characteristic of noise so, enters step 507, otherwise enters step 508.Number M is value rule of thumb, gets M=10 among this embodiment.

Step 507, the value of toneSNRbias is increased progressively h, h is value rule of thumb, gets h=2 among this embodiment;

Step 508, keep value constant of toneSNRbias.

Draw the value of toneSNRbias according to above-mentioned steps after, the value of toneSNRbias is added in does comprehensive judgement on the VAD decision threshold, promptly the conclusive judgement thresholding is vad_thr+toneSNRbias, and then operates by existing follow-up flow process.

Need to prove, can the end value of toneSNRbias be limited, make it in rational mobility scale, this mobility scale can certain negative value and on the occasion of between, to avoid the excessive or too small situation of this value appearance.

It should be noted that also above-mentioned said preset value thr, M, N, h etc. are the parameter of value rule of thumb, and value can be set as the case may be flexibly.

Can find, distinguish useful signal and noise signal by continuous tone information in the present embodiment, all there is not tone if detected continuous multiple frames, think that then noisiness is obvious, just thresholding additional corrections value is increased with slower speed, thereby improve whole decision threshold, guarantee that noise is not mistaken for useful signal; When having detected continuous tone, think that then the useful signal characteristic is obvious, thresholding additional corrections value fast speeds is reduced, be not mistaken for noise to guarantee useful signal.Therefore according to the feature of input signal, whole decision threshold is carried out up or down adjustment, improved the accuracy of VAD judgement.

Foregoing describes the method that the embodiment of the invention is adjusted threshold value in detail, and corresponding, the embodiment of the invention provides a kind of detection module.

Seeing also Fig. 6, is embodiment of the invention detection module structural representation.

Detection module comprises: judging unit 601, first processing unit 602, second processing unit 603 and the 3rd processing unit 604.

Judging unit 601 is used to judge the characteristics of signals of the signal of reception.

First processing unit 602, be used for when described judging unit 601 is judged described characterization and gone out the garbage signal characteristic, further judge whether to occur continuous garbage signal, when described judging unit 601 is judged described characterization and gone out the useful signal characteristic, further judge whether to occur continuous useful signal.Described first processing unit 602 further comprises continuous garbage signal processing unit 6021 and continuous useful signal processing unit 6022.Garbage signal processing unit 6021 is used to judge whether continuous garbage signal to occur continuously, is specially: whether the number of times that occurs the garbage signal characteristic more continuously surpasses first is set number, if then judge continuous garbage signal to occur; Useful signal processing unit 6022 is used to judge whether continuous useful signal to occur continuously, is specially: whether the number of times that has more continuously with characteristics of signals surpasses the second setting number, if then judge continuous useful signal to occur.

Second processing unit 603 is used for increasing thresholding additional corrections value when continuous garbage signal appears in 602 judgements of described first processing unit, when continuous useful signal appears in 602 judgements of described first processing unit, reduces thresholding additional corrections value.

The 3rd processing unit 604 is used to determine the threshold value that compares, with the former threshold value that is useful on comparison add increase in the above second processing unit 603 or reduce after thresholding additional corrections value as the threshold value that compares.

Detection module further comprises: characteristic acquiring unit 605.

Characteristic acquiring unit 605 is used for obtaining from the signal that receives the parameter of characterization signal characteristic.Characteristic acquiring unit 605 extracts the characteristic parameter that can characterize useful signal and garbage signal characteristic according to received signal.Characteristic parameter can be that entropy, tone or other can be represented the parameter of useful signal and garbage signal feature.Described judging unit 601 judges that the characteristics of signals of the signal that receives specifically is to judge by described characteristic acquiring unit 605 obtains the characterization signal characteristic from the signal that receives parameter.

Described judging unit 601 comprises first judging unit 6011 and second judging unit 6012.When the parameter of obtaining the characterization signal characteristic when characteristic acquiring unit 605 from the signal that receives is the entropy of signal, first judging unit 6011 compares the value and the preset value of the entropy of described signal, if greater than preset value, then characterize out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.Further, first judging unit 6011 will can also carry out smoothing processing with the entropy of described signal before the value of the entropy of described signal and the preset value comparison, and is corresponding, and the entropy and the preset value that will carry out the signal after the smoothing processing compare.

When the parameter of obtaining the characterization signal characteristic when characteristic acquiring unit 601 from the signal that receives is the tone of signal, second judging unit 6012 judges whether to contain the tone of signal according to the pitch marks that has produced, if the pitch marks of described generation does not contain the sign of the tone of signal for expression, then characterize out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.

In addition, described second processing unit 603 is judged at described first processing unit 602 and continuous garbage signal do not occurred, perhaps judge when continuous useful signal not occurring, keep thresholding additional corrections value constant, be specially: described second processing unit 603 compares the number of times that occurs the garbage signal characteristic continuously at described first processing unit 602 and surpasses the first setting number, the number of times that perhaps compares the non-garbage signal characteristic of continuous appearance surpasses second when setting number, keeps thresholding additional corrections value constant.Be worth when constant when described second processing unit 603 keeps the thresholding additional corrections, the 3rd processing unit 604 adds the thresholding additional corrections value that remains unchanged in the above second processing unit 603 as the threshold value that compares with the former threshold value that is useful on comparison.

Detection module further comprises: occur garbage signal counter 606 continuously and have continuously using event counter 607.

Occur garbage signal counter 606 continuously, be used to calculate the number of times that the garbage signal characteristic occurs.Have continuously with event counter 607, be used to calculate existing number of times with characteristics of signals.

After described judging unit 601 is judged the garbage signal characteristic that described parameter characterization goes out signal, describedly occur the number of times that the garbage signal characteristic will appear in garbage signal counter 606 continuously and add 1, described having continuously with event counter 607 will have number of times zero clearing with characteristics of signals; After described judging unit 601 is judged the useful signal characteristic that described parameter characterization goes out signal, the described number of times that will have with characteristics of signals with event counter 607 that has continuously adds 1, describedly occurs the number of times zero clearing that the garbage signal characteristic will appear in garbage signal counter 606 continuously.Like this, described first processing unit 602 just can be occurred the garbage signal characteristic continuously or be had number of times with characteristics of signals continuously according to occurring garbage signal counter 606 continuously and having count value with event counter 607 continuously.

In sum, because embodiment of the invention technical scheme is the situation that continues appearance according to different characteristic between useful signal and the garbage signal, judge whether to occur continuous useful signal or garbage signal, thereby self-adaptation increases or reduces thresholding additional corrections value, on the basis of original judgment thresholding suitable adjustment whole decision threshold, thereby solved in the prior art a large amount of signal misjudged problems under ground unrest situation that continues and lasting useful signal situation, improved the accuracy of signal decision.

Further, the characteristic parameter that the embodiment of institute of the present invention can characterize useful signal and garbage signal characteristic can be the entropy of signal, also can be tone, and perhaps other can represent the parameter of useful signal and garbage signal feature.

More than a kind of method and detection module of adjusting threshold value that the embodiment of the invention provided is described in detail, for one of ordinary skill in the art, thought according to the embodiment of the invention, part in specific embodiments and applications all can change, in sum, this description should not be construed as limitation of the present invention.

Claims (15)

1, a kind of method of adjusting threshold value is characterized in that, comprising:

Judge the characteristics of signals of the signal that receives;

If judging described characterization goes out the garbage signal characteristic, further judge whether to occur continuous garbage signal, if judging described characterization goes out the useful signal characteristic, further judge whether to occur continuous useful signal;

If continuous garbage signal then increases thresholding additional corrections value, if continuous useful signal then reduces thresholding additional corrections value;

The former threshold value that is useful on comparison added the above increase or reduce after thresholding additional corrections value as the threshold value that compares.

2, the method for adjustment threshold value according to claim 1 is characterized in that:

Describedly judge whether that continuous garbage signal occurring is specially: whether the number of times that occurs the garbage signal characteristic more continuously surpasses first is set number, if then judge continuous garbage signal to occur;

Describedly judge whether that continuous useful signal occurring is specially: have number of times with characteristics of signals more continuously and whether surpass second and set number, if then judge continuous useful signal to occur.

3, the method for adjustment threshold value according to claim 1 and 2 is characterized in that:

The characteristics of signals of the signal that described judgement receives specifically is to judge by the parameter of obtaining the characterization signal characteristic from the signal that receives.

4, the method for adjustment threshold value according to claim 3 is characterized in that:

The described parameter of obtaining the characterization signal characteristic from the signal that receives is the entropy of signal;

Accordingly, the entropy of the judging described signal garbage signal characteristic that characterizes out signal is specially:

With the value of the entropy of described signal and preset value relatively, if greater than preset value, then characterize out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.

5, the method for adjustment threshold value according to claim 4 is characterized in that:

Comprise that further the entropy with described signal carries out smoothing processing before the value of the entropy of described signal and the preset value comparison, corresponding, the entropy and the preset value that will carry out the signal after the smoothing processing compare.

6, the method for adjustment threshold value according to claim 3 is characterized in that:

The described parameter of obtaining the characterization signal characteristic from the signal that receives is the tone of signal;

Accordingly, the tone of the judging described signal garbage signal characteristic that characterizes out signal is specially:

The pitch marks that produces according to scrambler judges whether to contain the tone of signal, if the pitch marks of described generation then characterizes out the garbage signal characteristic of signal, otherwise characterizes out the useful signal characteristic of signal for expression does not contain the sign of the tone of signal.

7, the method for adjustment threshold value according to claim 3 is characterized in that:

Whether the described number of times that occurs the garbage signal characteristic more continuously surpasses first further comprises before setting number: after judging the garbage signal characteristic that described parameter characterization goes out signal, to occur the garbage signal counter continuously and add 1, will have continuously and use the event counter zero clearing;

Whether described having more continuously surpasses second with the number of times of characteristics of signals and further comprises before setting number: after judging the useful signal characteristic that described parameter characterization goes out signal, to have with event counter continuously and add 1, will occur the garbage signal counter O reset continuously.

8, the method for adjustment threshold value according to claim 1 is characterized in that:

If continuous garbage signal do not occur, when continuous useful signal perhaps not occurring, then keep thresholding additional corrections value constant, corresponding,

The former threshold value that is useful on comparison is added the above thresholding additional corrections value that remains unchanged as the threshold value that compares.

9, a kind of detection module is characterized in that, comprising:

Judging unit is used to judge the characteristics of signals of the signal of reception;

First processing unit, be used for when described judgment unit judges goes out described characterization and goes out the garbage signal characteristic, further judge whether to occur continuous garbage signal, when described judgment unit judges goes out described characterization and goes out the useful signal characteristic, further judge whether to occur continuous useful signal;

Second processing unit is used for increasing thresholding additional corrections value when continuous garbage signal appears in described first processing unit judgement, when continuous useful signal appears in described first processing unit judgement, reduces thresholding additional corrections value;

The 3rd processing unit is used to determine the threshold value that compares, with the former threshold value that is useful on comparison add increase in the above second processing unit or reduce after thresholding additional corrections value as the threshold value that compares.

10, detection module according to claim 9 is characterized in that, described detection module further comprises:

The characteristic acquiring unit is used for obtaining from the signal that receives the parameter of characterization signal characteristic;

The characteristics of signals of the signal that described judgment unit judges receives specifically is to judge by described characteristic acquiring unit obtains the characterization signal characteristic from the signal that receives parameter.

11, detection module according to claim 10 is characterized in that:

Described characteristic acquiring unit obtains the characterization signal characteristic from the signal that receives parameter is the entropy of signal;

Described judging unit comprises first judging unit, is used for value and preset value comparison with the entropy of described signal, if greater than preset value, then characterizes out the garbage signal characteristic of signal, otherwise characterizes out the useful signal characteristic of signal.

12, detection module according to claim 10 is characterized in that:

Described characteristic acquiring unit obtains the characterization signal characteristic from the signal that receives parameter is the tone of signal;

Described judging unit comprises second judging unit, be used for judging whether to contain the tone of signal according to the pitch marks that produced, if the pitch marks of described generation does not contain the sign of the tone of signal for expression, then characterize out the garbage signal characteristic of signal, otherwise characterize out the useful signal characteristic of signal.

13, according to each described detection module of claim 9 to 12, it is characterized in that described detection module further comprises:

Occur the garbage signal counter continuously, be used to calculate the number of times that the garbage signal characteristic occurs;

Have continuously and use event counter, be used to calculate existing number of times with characteristics of signals;

After described judgment unit judges goes out the garbage signal characteristic that described parameter characterization goes out signal, describedly occur the number of times that the garbage signal characteristic will appear in the garbage signal counter continuously and add 1, described having continuously with event counter will have number of times zero clearing with characteristics of signals;

After described judgment unit judges goes out the useful signal characteristic that described parameter characterization goes out signal, the described number of times that will have with characteristics of signals with event counter that has continuously adds 1, describedly occurs the number of times zero clearing that the garbage signal characteristic will appear in the garbage signal counter continuously.

14, detection module according to claim 9 is characterized in that, described first processing unit further comprises:

The garbage signal processing unit is used for occurring more continuously the number of times of garbage signal characteristic whether above the first setting number, if then judge the garbage signal that appearance is continuous continuously;

The useful signal processing unit is used for having more continuously the number of times of using characteristics of signals and whether surpasses the second setting number continuously, if then judge continuous useful signal to occur.

15, detection module according to claim 9 is characterized in that:

Described second processing unit is judged at described first processing unit and continuous garbage signal do not occurred, perhaps judges when continuous useful signal not occurring, keeps thresholding additional corrections value constant, corresponding,

The 3rd processing unit adds the thresholding additional corrections value that remains unchanged in the above second processing unit as the threshold value that compares with the former threshold value that is useful on comparison.

Images