CN101820302A - Device and method for canceling echo - Google Patents
Device and method for canceling echo Download PDFInfo
- Publication number
- CN101820302A CN101820302A CN200910105667A CN200910105667A CN101820302A CN 101820302 A CN101820302 A CN 101820302A CN 200910105667 A CN200910105667 A CN 200910105667A CN 200910105667 A CN200910105667 A CN 200910105667A CN 101820302 A CN101820302 A CN 101820302A
- Authority
- CN
- China
- Prior art keywords
- signal
- far
- pitch period
- input signal
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention discloses a device for canceling echo, which comprises a near-end cache module, a near-end voice detection module, a far-end voice detection module, a near-end pitch period module, a far-end pitch period module and a signal separation module. The device can analyze independent components according to a near-end input signal, calculates each signal source and a pitch period thereof when the analyzing result of the independent components is converged, compares the pitch period of each signal source with a near-end pitch period and a far-end pitch period to obtain an acoustic echo signal, further obtains the near-end input signal after the echo is cancelled and outputs the near-end input signal to a far end so as to effectively cancel the acoustic echo in real time.
Description
Technical field
The present invention relates to field of voice signal, be specifically related to a kind of echo cancelling device and method.
Background technology
Along with the arrival of information age, the communication mode that people day by day rely on develops from the direction of early stage single speech communication to multiple business, multiple network integrated communication.In the various types of communication business, every needs use the occasion of voice playing equipment and voice acquisition equipment simultaneously, and as videoconference, video conference, Internet phone-calling etc., echo has had influence on calling quality to a certain extent.Echo can be divided into electric echo and acoustic echo, and wherein electric echo mainly is not cause owing to the data transaction that exists in the communication system matches.
Acoustic echo is owing to the coupling of the voice between voice playing equipment and the voice acquisition equipment forms, and concrete, owing to adopted full-duplex channel, the loud speaker of near-end and far-end and microphone carry out work simultaneously.Near end input signal is exported by far-end loudspeaker, and the signal of this loud speaker output will be picked up the loud speaker of directly passing near-end back by the microphone of far-end, causes the microphone of near-end to capture this signal, thereby has produced acoustic echo.
In order to increase the stability of full duplex communication system, improve communication quality, the echo cancelling device with adaptive-filtering function commonly used at present is set to solve echo problem in the system relevant position.
So-called adaptive-filtering is exactly the result who utilizes the acquired filter parameter of previous moment, regulates the filter parameter of current time automatically, adapting to the input signal of current time, thereby realizes optimization filtering.Echo cancelling device with adaptive-filtering function is according to the characteristic parameter of the echo path that estimates, and to produce an analog echo signal, deducts above-mentioned echo signal from the signal that receives, thereby realizes echo elimination.
But in this Echo Canceller device, owing to lack accurate control, and when the bigger situation of ambient noise even can cause can't steady operation, can not eliminate echo effectively, to such an extent as to artificially introduce noise to sef-adapting filter.And under different application scenarios, need constantly debugging external parameter step factor, to find out a more rational step factor.If the speaker walks about or a plurality of people talks simultaneously (promptly existing under the more serious acoustic echo situation), can influence the operating state of sef-adapting filter.
Summary of the invention
The existing adaptive-filtering echo eliminating device echo of the problem to be solved in the present invention is eliminated the problem of poor effect, thereby at acoustic echo, proposes a kind of less echo eliminating device that is affected by the external environment that provides.
The present invention is achieved in that
A kind of echo cancelling device comprises:
The near-end cache module is used for receiving and preserving near end input signal, and near end input signal is outputed to signal separation module;
The near-end speech detection module is used to receive near end input signal, when judging this signal and being voice signal, sends near-end fundamental tone triggering signal to near-end pitch period module;
The far-end speech detection module is used to receive remote end input signal, when judging this signal and being voice signal, sends far-end fundamental tone triggering signal to far-end pitch period module;
Near-end pitch period module is used to receive the near-end fundamental tone triggering signal of near end input signal and the transmission of described near-end speech detection module, and according near end input signal generation near-end speech pitch period, exports signal separation module to;
Far-end pitch period module is used to receive the far-end fundamental tone triggering signal of remote end input signal and the transmission of described far-end speech detection module, and according to remote end input signal generation far-end speech pitch period, exports signal separation module to;
Signal separation module, the near end input signal, the near-end pitch period module that are used to receive the transmission of near-end cache module send the far-end speech pitch period of near-end speech pitch period, the transmission of far-end pitch period module, described signal separation module is carried out independent component analysis according near end input signal, when the independent component analysis result restrains, calculate each signal source and pitch period thereof; Described signal separation module compares the pitch period of described each signal source with described near-end speech pitch period and far-end speech pitch period, obtain echo signal and also export adder to; Adder is calculated the difference of near end input signal and echo signal, and the near end signal that obtains after echo is eliminated is also exported this signal.
The present invention has further proposed a kind of echo cancel method, and wherein, this method comprises:
A, near-end cache module are preserved near end input signal, and simultaneously, the near-end speech detection module is preserved near end input signal and judged whether current this signal is voice signal, if voice signal execution in step B; The far-end speech detection module is preserved remote end input signal and is judged whether current this signal is voice signal simultaneously, if voice signal execution in step C;
B, near-end pitch period module receive near end input signal and generate the near-end speech pitch period and send to signal separation module according near end input signal;
C, far-end pitch period module receive remote end input signal and generate the far-end speech pitch period and send to signal separation module according to remote end input signal;
D, signal separation module are carried out independent component analysis according to the near end input signal of preserving, and when the independent component analysis result restrains, calculate each signal source and pitch period thereof; And according to the near-end speech pitch period and the far-end speech pitch period that receive, the pitch period of described each signal source is compared with near-end speech pitch period and far-end speech pitch period, obtain echo signal, the near end signal that calculates after acoustic echo is eliminated exports far-end to.
Compared with prior art, echo cancelling device provided by the invention is calculated each signal source according to the near end input signal analysis meter, then by relatively judging fundamental tone, isolate echo signal, and then elimination echo, echo cancelling device of the present invention can carry out above-mentioned echo cancellation in real time, and it is little to be affected by the external environment, thereby effectively eliminates echo.
Description of drawings
Fig. 1 uses the present invention to realize the schematic diagram that acoustic echo is eliminated in the communication network;
Fig. 2 is the theory diagram that the present invention realizes the specific embodiment one that acoustic echo is eliminated;
Fig. 3 is the theory diagram that the present invention realizes the specific embodiment two that acoustic echo is eliminated;
Fig. 4 is the theory diagram that the present invention realizes the specific embodiment three that acoustic echo is eliminated;
Fig. 5 is the theory diagram that the present invention realizes the specific embodiment four that acoustic echo is eliminated;
Fig. 6 is the theory diagram that the present invention realizes the specific embodiment five that acoustic echo is eliminated;
Fig. 7 is the theory diagram that the present invention realizes the specific embodiment six that acoustic echo is eliminated.
Embodiment
For making purpose of the present invention, technical scheme and advantage clearer, below with reference to the accompanying drawing embodiment that develops simultaneously, the present invention is described in more detail.
Fig. 1 carries out the schematic diagram that the near-end acoustic echo is eliminated and the far-end acoustic echo is eliminated respectively for both-end in the communication network uses two echo cancelling devices of the present invention, and the signal with far-end acoustic echo of near-end input is exported through second echo cancelling device; The signal with near-end acoustic echo of far-end input is exported through first echo cancelling device, thereby makes that under the full duplex working condition, both-end all has voice quality preferably.
Fig. 2 is the theory diagram that the present invention realizes the specific embodiment one that acoustic echo is eliminated, and Figure 2 shows that the theory diagram of Fig. 1 frame of broken lines part.Sin among the figure represents that near-end will send to the initialize signal of far-end, it is near end input signal, this near end input signal Sin comprises one or more in voice signal, ambient noise signal, the acoustic echo signal, and described near end input signal is observed via a plurality of observation path; Sout represents that near end input signal is treated, outputs to the signal of far-end, the near end signal after promptly ambient noise signal and acoustic echo are eliminated.Rin represents that far-end will send to the initialize signal of near-end, i.e. remote end input signal; Rout represents that remote end input signal is transferred to the signal of near-end.
As shown in Figure 2, echo cancelling device mainly comprises:
Near-end cache module 1 is used for receiving and preserving Sin, and Sin is outputed to signal separation module 6;
Near-end speech detection module 2 is used to receive Sin, when judging this signal and being voice signal, sends near-end fundamental tone triggering signals to near-end pitch period module 4;
Far-end speech detection module 3 is used to receive Rin, when judging this signal and being voice signal, sends far-end fundamental tone triggering signals to far-end pitch period module 5;
Near-end pitch period module 4 is used to receive near-end fundamental tone triggering signal and the Sin that near-end speech detection module 2 sends, and described near-end pitch period module 4 generates the near-end speech pitch period and sends to signal separation module 6 according to Sin;
Far-end pitch period module 5 is used to receive far-end fundamental tone triggering signal and the Rin that far-end speech detection module 3 sends, and described far-end pitch period module 5 generates the far-end speech pitch period and sends to signal separation module 6 according to Rin;
Signal separation module 6, the near end input signal, the near-end pitch period module 4 that are used to receive 1 transmission of near-end cache module send the far-end speech pitch period of near-end speech pitch periods, 5 transmissions of far-end pitch period module, described signal separation module 6 is carried out independent component analysis according to Sin, when the independent component analysis result restrains, calculate each signal source and pitch period thereof; Described signal separation module 6 compares the pitch period of described each signal source with described near-end speech pitch period and far-end speech pitch period, obtain echo signal re ' and also export adder to; Adder is calculated the difference of Sin and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
Above-mentioned echo signal re ' comprises acoustic echo signal and ambient noise signal.
Echo cancelling device provided by the invention is calculated each signal source according to the near end input signal analysis meter, then by relatively judging fundamental tone, isolate the acoustic echo signal, and then elimination echo, echo cancelling device of the present invention can carry out above-mentioned echo cancellation in real time, it is little to be affected by the external environment, thereby effectively eliminates acoustic echo.
Based on specific embodiment one, echo cancelling device of the present invention also provides embodiment two.Fig. 3 is the theory diagram of specific embodiment two, and as shown in Figure 3, this embodiment has comprised the whole modules of Fig. 2, and:
Signal separation module comprises:
Albefaction unit 61: be used for the near end input signal from a plurality of observation path is formed hybrid matrix, and described hybrid matrix is carried out albefaction, obtain whitened signal and export to computing unit 62;
Computing unit 62: be used to receive whitened signal and the Sin that described albefaction unit 61 sends, and carry out iterative computation according to described whitened signal, obtain separation matrix during convergence, described computing unit 62 calculates each signal source and exports extraction unit 63 to according to Sin and separation matrix;
Extraction unit 63: each signal source, the near-end pitch period module 4 that are used to receive the computing unit transmission send the far-end speech pitch period of near-end speech pitch periods, 5 transmissions of far-end pitch period module, described extraction unit 63 compares the pitch period of each signal source with described near-end speech pitch period and far-end speech pitch period, obtain echo signal and also export adder to; Adder is calculated the difference of Sin and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
Fig. 4 is the theory diagram of echo cancelling device embodiment three of the present invention, compares with embodiment one, and this embodiment has comprised the whole modules of Fig. 2, and:
The near-end speech detection module comprises near-end energy calculation unit 21 and near-end speech judging unit 22.
Described near-end energy calculation unit 21, be used to receive near end signal Sin, calculate the short time ENERGY E Sin of near end input signal, and according to the maximum of the energy of the Sin between quiet period, with the numerical value that obtains after this maximum multiplication by constants 1.2 as near-end speech threshold values ETs, and with near-end speech threshold values ETs and the ENERGY E Sin of the Sin when formally conversing output to near-end speech judging unit 22;
Described near-end speech judging unit 22, be used to receive near end input signal short time ENERGY E Sin and near-end speech threshold value ETs, and both are compared, as near end input signal short time ENERGY E Sin during greater than near-end speech threshold value ETs, the judgement near end input signal is a voice signal; When near end input signal is voice signal, send near-end fundamental tone triggering signal to near-end pitch period module 4.
Wherein,
The short-time energy computing formula of near end input signal is:
Certainly near-end speech threshold value ETs can be a preset value, and for example, can rule of thumb get it is 0.001.
The far-end speech detection module comprises far-end energy calculation unit 31 and far-end speech judging unit 32.
Described far-end energy calculation unit 31, be used to receive remote signaling Rin, calculate the short time ENERGY E Rin of remote end input signal, and according to the maximum of the energy of the Rin between quiet period, with the numerical value that obtains after this maximum multiplication by constants 1.2 as far-end speech threshold values ETr, and with ETr and the ENERGY E Rin of the Rin when formally conversing output to far-end speech judging unit 32;
Described far-end speech judging unit 32, be used to receive remote end input signal short time ENERGY E Rin and far-end speech threshold value ETr, and both are compared, as remote end input signal short time ENERGY E Rin during greater than far-end speech threshold value ETr, the judgement remote end input signal is a voice signal; When remote end input signal is voice signal, send far-end fundamental tone triggering signal to far-end pitch period module 5.
Wherein,
The short-time energy computing formula of remote end input signal is:
Certainly far-end speech threshold value ETs can be a preset value, and for example rule of thumb getting it is 0.001.
Fig. 5 is the theory diagram of echo cancelling device embodiment four of the present invention, compares with embodiment one, and this embodiment has comprised the whole modules of Fig. 2, and:
Near-end pitch period module comprises:
Near-end low-pass filter unit 41 is used to receive near-end fundamental tone triggering signal and the Sin that near-end speech detection module 2 sends, and Sin is carried out Filtering Processing.The influence that described near-end low-pass filter unit 41 can filter away high frequency noise disturbs also can be played weaken that the multiple harmonic component plays the effect that does not weaken near end input signal fundamental frequency information to the influence of first formant in the near end input signal frequency spectrum simultaneously;
Near-end Fourier transformation unit 42 is used to generate the spectrum information of near end input signal and exports near-end maximum likelihood decision unit 44 to;
Near-end linear prediction unit 43 is used for Sin and carries out the linear prediction processing, and the frequency spectrum that forms the sound channel impulse response of Sin also exports near-end maximum likelihood decision unit 44 to;
Near-end maximum likelihood decision unit 44, be used to receive the spectrum information of near-end Fourier transformation unit 42 transmissions and the frequency spectrum of the sound channel impulse response that near-end linear prediction unit 43 sends, and generate the near-end speech pitch period and send to signal separation module 4 according to both.
Far-end pitch period module comprises:
Far-end low-pass filter unit 51 is used to receive far-end fundamental tone triggering signal and the Rin that far-end speech detection module 3 sends, and Rin is carried out Filtering Processing.The influence that described far-end low-pass filter unit 51 can filter away high frequency noise disturbs also can be played weaken that the multiple harmonic component plays the effect that does not weaken remote end input signal fundamental frequency information to the influence of first formant in the remote end input signal frequency spectrum simultaneously;
Far-end Fourier transformation unit 52 is used to generate the spectrum information of remote end input signal and exports far-end maximum likelihood decision unit 54 to;
Far-end linear prediction unit 53 is used for Rin and carries out the linear prediction processing, and the spectrum information that forms the sound channel impulse response of Rin also exports far-end maximum likelihood decision unit 54 to;
Far-end maximum likelihood decision unit 54, be used to receive the spectrum information of far-end Fourier transformation unit 52 transmissions and the spectrum information of the sound channel impulse response that far-end linear prediction unit 53 sends, and generate the far-end speech pitch period and send to signal separation module 4 according to both.
Each unit in the far-end pitch period module can adopt with near-end pitch period module in the identical data processing method in each unit, be example with near-end pitch period module below, the concrete processing method of its each unit is described:
A kind of selection of near-end low-pass filter unit 41 can be the 5 rank low pass filters that cut-off frequency is 800Hz, and Sin is carried out Filtering Processing.
Near-end Fourier change unit 42 receives through the signal after the low pass filter unit processing, searches for the short-term spectrum of near end input signal in the frame
The formant of the first corresponding maximum, and corresponding peaks is converted to the time domain peak value, be designated as x
fWith [x
f-1
f+ 1] as the value according to a preliminary estimate of pitch period, carry out periodic extension then and obtain glottal excitation, add and carry out spectrum analysis by Fourier transform module behind the Hamming window and obtain
This frequency spectrum is sent into maximum likelihood decision unit 44;
Near-end linear prediction unit 43 receives Sin and carries out the linear prediction processing, the sound channel impulse response that calculates, and make corresponding frequency spectrum
Send into the maximum likelihood decision unit.Because have correlation between the voice sampling point, then the sample value in available past is predicted the present or following sample value, i.e. several voice sampling of the sampling of voice enough past of energy or their linear combination approach.
A in the formula
iBe past voice sampling constantly S
In(n-i) weight coefficient, p (desirable p=10) is a prediction order.Then the difference of actual signal and prediction signal is prediction residual
According to least mean-square error LMS criterion, if make E[|e (n) |
2] minimum, then can determine one group of unique linear predictor coefficient a
i(i=1,2 ... p).Determined each predictive coefficient a
iAfter, can obtain the frequency spectrum of its frequency response
And export to near-end maximum likelihood decision unit 44, wherein
Described near-end maximum likelihood decision unit 44 is according to the frequency spectrum that receives
With
Restructural goes out the short-term spectrum of original input signal
Will
Carry out similarity relatively, calculating formula is as follows:
At both mean square error ε
) value of short-term spectrum correspondence at minimum value place is revised pitch period.Described near-end maximum likelihood decision unit 44 exports the near-end speech pitch period that obtains to signal processing module 4, to carry out the relevant treatment of fundamental tone coupling.
Fig. 6 is the theory diagram of echo cancelling device embodiment five of the present invention, compares with embodiment one, and this embodiment has comprised the whole modules of Fig. 2, comprises that further a control module 7 and receives the output module 8 of the control signal of control module 7 transmissions,
Described near-end speech detection module 2 receives near end input signal, when judging this signal and being voice signal, sends the near-end speech triggering signals to control module 7;
Described far-end speech detection module 3 receives remote end input signal, when judging this signal and being voice signal, sends the far-end speech triggering signals to control module 7;
Described control module 7 is when receiving only the near-end speech triggering signal that near-end speech detection module 2 sends or do not receive any speech trigger information, and the output control signal is given output module 8; Described output module 8 receives above-mentioned control signal, and with near end input signal to far-end;
When receiving only the far-end speech triggering signal that far-end speech detection module 3 sends, when perhaps receiving the far-end speech triggering signal that near-end speech triggering signal that near-end speech detection module 2 sends and far-end speech detection module 3 send simultaneously, control described signal separation module 6 and receive the near end input signal that near-end cache module 1 sends, near-end pitch period unit sends the near-end speech pitch period, the far-end speech pitch period that far-end pitch period module sends, described signal separation module 6 is carried out independent component analysis according near end input signal, when the independent component analysis result restrains, calculate each signal source and pitch period thereof; Described signal separation module 6 compares the pitch period of described each signal source with described near-end speech pitch period and far-end speech pitch period, obtain the acoustic echo signal and also export adder to; Adder is calculated the difference of near end input signal and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
On the basis based on the foregoing description five, echo cancelling device of the present invention provides a specific embodiment six, compares with embodiment four, and this embodiment has comprised whole modules of embodiment four, wherein,
Described output module further comprises a comfort noise generation unit and a signal output unit,
Described control module is when receiving only the near-end speech triggering signal that the near-end speech detection module sends or do not receive any speech trigger information, and output one controls signal to the comfort noise generation unit;
Described comfort noise generation unit is used to receive the control signal that described control module sends, and generates the comfort noise signal output signal output unit of certain level;
Signal output unit is with near end input signal and the stack of comfort noise signal, and the signal after will superposeing exports far-end to.
Obviously, echo cancelling device with above-mentioned comfort noise generation unit, can all not have under the situation of voice signal by both-end, the comfort noise signal that certain level is provided is to the far-end correspondent, thereby can be effective when avoiding both-end all not have voice signal, the user thinks easily that circuit has interrupted by mistake or the generation of phenomenon such as apparatus failure.
Below be the specific implementation method of the specific embodiment one of realization echo provided by the invention elimination:
S01: the near-end cache module is preserved Sin, and simultaneously, the near-end speech detection module is preserved Sin and judged whether current this signal is voice signal, if voice signal execution in step S02; The far-end speech detection module is preserved Rin and is judged whether current this signal is voice signal simultaneously, if voice signal execution in step S03;
S02: near-end pitch period module receives Sin and generates the near-end speech pitch period and send to signal separation module according to Sin;
S03: far-end pitch period module receives Rin and generates the far-end speech pitch period and send to signal separation module according to Rin;
S04: signal separation module is carried out independent component analysis according to the Sin that preserves, and when the independent component analysis result restrains, calculates each signal source and pitch period thereof; And according to the near-end speech pitch period and the far-end speech pitch period that receive, the pitch period of described each signal source is compared with near-end speech pitch period and far-end speech pitch period, obtain the acoustic echo signal, the near end signal that calculates after acoustic echo is eliminated exports far-end to.
Signal separation module carries out independently being divided into analysis in real time, when analysis result is restrained, even if calculate separation matrix, thereby obtains each signal source.
When near-end and remote end input signal all are under the non-speech audio situation, Sin only is the near-end ambient noise signal; When only having Sin to be voice signal, Sin comprises near-end voice signals and near-end ambient noise signal; When only having Rin to be voice signal, Rin comprises far-end speech signal and distal environment noise signal, and Sin comprises the echo signal of near-end ambient noise signal and Rin generation; When the both-end input signal all was voice signal, Rin comprised echo signal, far-end input speech signal and the distal environment noise signal that near end input signal Sin produces.
By top analysis as can be known, when near-end and remote end input signal all are that the situation of voice is the most complicated, other situations can be regarded the simplification of above-mentioned situation as, all are that voice signal is an example with the both-end input signal below, specify the step that signal separation module calculates each signal source:
In order to express easily, the present invention is reduced near end input signal
S(n)＝[S
1(n)S
2(n)S
3(n)]
S wherein
1(n), S
2(n), S
3(n) three near end input signal that microphone receives respectively are respectively near-end voice signals, the echo signal that is produced by remote end input signal, the mixed signal that is superimposed as in various degree of near-end ambient noise signal.
SP01, whitened signal hybrid matrix X (n) obtain albefaction matrix Y (n), Y (n)=UX (n).Whitening approach is for to carry out characteristic value decomposition to the covariance of signal hybrid matrix X (n), makes R
x=V Λ V
TMake U=Λ
-1/2V
T, then obtain whitened signal Y (n)=Λ
-1/2V
TX (n).
SP02, signal separation module are to estimate a separation matrix W on the one hand, make
Make its each component approach component among the S (n).The iterative step of algorithm is as follows in one frame.
1) makes i=1.
2) initialization matrix vector w (0), and make k=1.
3) make w
i(k)=E[Y
i(w (k-1)
TY
i)
3]-3w
i(k-1)..
4) order
In order to ensure estimate a different isolated component at every turn, need in circulation, add a rectangular projection, obtain
5) if | w
i(k)
Tw
i(k-1) | converge on 1, then stop iteration, output w
i(k), otherwise make k=k+1, return 3) step, continue iteration.Up to obtaining separation matrix W, this W can be designated as:
SP03, basis
Calculate each signal source, calculate the fundamental tone of each signal source simultaneously.
Both-end all has the situation of voice signal the most complicated, and other situations are simplification of above-mentioned situation, and the concrete steps of signal extraction are example when all voice signal being arranged with both-end below, explain signal extraction:
SP04, the fundamental tone of each signal source and far-end input voice fundamental cycle are compared, can determine to be in the signal source voice signal of remote end input signal, remaining as can be known two paths of signals is acoustic echo signal and the distal environment noise signal that near-end produces simultaneously, is the signal that needs elimination.
Compared with prior art, echo cancel method provided by the invention is calculated each signal source according to the near end input signal analysis meter, then by relatively judging fundamental tone, isolate echo signal, and then elimination echo, echo cancel method of the present invention can carry out above-mentioned echo cancellation in real time, and it is little to be affected by the external environment, thereby effectively eliminates echo.
The above is preferred embodiment of the present invention only, is not to be used to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any modification of being done, be equal to and replace and improvement etc., all should be included within protection scope of the present invention.
Claims (9)
1. an echo cancelling device is characterized in that, comprises
The near-end cache module is used for receiving and preserving near end input signal, and near end input signal is outputed to signal separation module;
The near-end speech detection module is used to receive near end input signal, when judging this signal and being voice signal, sends near-end fundamental tone triggering signal to near-end pitch period module;
The far-end speech detection module is used to receive remote end input signal, when judging this signal and being voice signal, sends far-end fundamental tone triggering signal to far-end pitch period module;
Near-end pitch period module is used to receive the near-end fundamental tone triggering signal of near end input signal and the transmission of near-end speech detection module, and according near end input signal generation near-end speech pitch period, exports signal separation module to;
Far-end pitch period module is used to receive the far-end fundamental tone triggering signal of remote end input signal and the transmission of far-end speech detection module, and according to remote end input signal generation far-end speech pitch period, exports signal separation module to;
Signal separation module, the near end input signal, the near-end pitch period module that are used to receive the transmission of near-end cache module send the far-end speech pitch period of near-end speech pitch period, the transmission of far-end pitch period module, described signal separation module is carried out independent component analysis according near end input signal, when the independent component analysis result restrains, calculate each signal source and pitch period thereof; Described signal separation module compares the pitch period of described each signal source with described near-end speech pitch period and far-end speech pitch period, obtain echo signal and also export adder to; Adder is calculated the difference of near end input signal and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
2. echo cancelling device according to claim 1 is characterized in that, described signal separation module comprises:
The albefaction unit is used to receive described near end input signal and forms hybrid matrix, and described hybrid matrix is carried out albefaction, obtains whitened signal and exports to computing unit;
Computing unit, be used to receive whitened signal and the near end input signal that described albefaction unit sends, carry out iterative computation, obtain separation matrix during the iterative computation convergence according to described whitened signal, and calculate each signal source according to described near end input signal and separation matrix, export extraction unit to;
Extraction unit, each signal source, the near-end pitch period module that are used to receive the computing unit transmission send the far-end speech pitch period of near-end speech pitch period, the transmission of far-end pitch period module, the pitch period of each signal source is compared with described near-end speech pitch period and far-end speech pitch period, obtain echo signal and also export adder to; Adder is calculated the difference of near end input signal and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
3. echo cancelling device according to claim 1 is characterized in that, described near-end speech detection module comprises:
Described near-end energy calculation unit, be used to receive near end input signal, calculate the short time energy of this signal, and determine the near-end speech threshold values according to the energy of this signal of quiet period, and the energy of the near end input signal during with this near-end speech threshold values and formal conversation outputs to the near-end speech judging unit;
Described near-end speech judging unit, the energy of the near end input signal when being used to receive near-end speech threshold values and formal conversation, and both are compared judge whether near end input signal is voice signal; When near end input signal is voice signal, send near-end fundamental tone triggering signal to near-end pitch period module.
4. echo cancelling device according to claim 1 is characterized in that, described far-end speech detection module comprises:
Described far-end energy calculation unit, be used to receive remote end input signal, calculate the short time energy of this signal, and determine the far-end speech threshold values according to the energy of this signal of quiet period, and the energy of the remote end input signal during with this far-end speech threshold values and formal conversation outputs to the far-end speech judging unit;
Described far-end speech judging unit, the energy of the remote end input signal when being used to receive far-end speech threshold values and formal conversation, and both are compared judge whether remote end input signal is voice signal; When remote end input signal is voice signal, send far-end fundamental tone triggering signal to far-end pitch period module.
5. according to the described echo cancelling device of claim 1, it is characterized in that described near-end pitch period module comprises:
The near-end low-pass filter unit is used to receive the near-end fundamental tone triggering signal that near end input signal and near-end speech detection module send, and near end input signal is carried out Filtering Processing and exported the near-end Fourier transformation unit to and the near-end linear prediction unit;
The near-end Fourier transformation unit is used to receive and according to the near end input signal that the near-end low-pass filter unit is handled, generates the spectrum information of near end input signal, exports near-end maximum likelihood decision unit to;
The near-end linear prediction unit is used to receive and according to the near end input signal that the near-end low-pass filter unit is handled, this signal is carried out linear prediction handle, and the frequency spectrum of the sound channel impulse response of formation near end input signal also exports near-end maximum likelihood decision unit to;
Near-end maximum likelihood decision unit is used to receive the spectrum information of near-end Fourier transformation unit transmission and the frequency spectrum of the sound channel impulse response that the near-end linear prediction unit sends, and generates the near-end speech pitch period and send to signal separation module according to both.
6. echo cancelling device according to claim 1 is characterized in that, described far-end pitch period module comprises:
The far-end low-pass filter unit is used to receive the far-end fundamental tone triggering signal that remote end input signal and far-end speech detection module send, and remote end input signal is carried out Filtering Processing and exported the far-end Fourier transformation unit to and the far-end linear prediction unit;
The far-end Fourier transformation unit is used to receive and according to the remote end input signal that the far-end low-pass filter unit is handled, generates the spectrum information of remote end input signal, exports far-end maximum likelihood decision unit to;
The far-end linear prediction unit is used to receive and according to the remote end input signal that the far-end low-pass filter unit is handled, this signal is carried out linear prediction handle, and the frequency spectrum of the sound channel impulse response of formation remote end input signal also exports far-end maximum likelihood decision unit to;
Far-end maximum likelihood decision unit is used to receive the spectrum information of far-end Fourier transformation unit transmission and the frequency spectrum of the sound channel impulse response that the far-end linear prediction unit sends, and generates the far-end speech pitch period and send to signal separation module according to both.
7. echo cancelling device according to claim 1 is characterized in that, comprises that also a control module and receives the output module of the control signal of control module transmission,
Described near-end speech detection module receives near end input signal, when judging this signal and being voice signal, sends the near-end speech triggering signal to control module;
Described far-end speech detection module receives remote end input signal, when judging this signal and being voice signal, sends the far-end speech triggering signal to control module;
Described control module is when receiving only the near-end speech triggering signal that the near-end speech detection module sends or do not receive any speech trigger information, and the output control signal is given output module; Described output module receives above-mentioned control signal, and with near end input signal to far-end;
When receiving only the far-end speech triggering signal that the far-end speech detection module sends, when perhaps receiving the far-end speech triggering signal that near-end speech triggering signal that the near-end speech detection module sends and far-end speech detection module send simultaneously, control described signal separation module and receive the near end input signal that the near-end cache module sends, near-end pitch period unit sends the near-end speech pitch period, the far-end speech pitch period that far-end pitch period module sends, described signal separation module is carried out independent component analysis according near end input signal, when the independent component analysis result restrains, calculate each signal source and pitch period thereof; Described signal separation module compares the pitch period of described each signal source with described near-end speech pitch period and far-end speech pitch period, obtain echo signal and also export adder to; Adder is calculated the difference of near end input signal and echo signal, obtains the near end signal after echo is eliminated and exports this signal to far-end.
8. echo cancelling device according to claim 7 is characterized in that, described output module advances to comprise a comfort noise generation unit and a signal output unit,
Described control module is when receiving only the near-end speech triggering signal that the near-end speech detection module sends or do not receive any speech trigger information, and output one controls signal to the comfort noise generation unit;
Described comfort noise generation unit is used to receive the control signal that described control module sends, and generates the comfort noise signal of certain level, output signal output unit;
Described signal output unit is with near end input signal and the stack of comfort noise signal, and the signal after will superposeing exports far-end to.
9. an echo cancel method is characterized in that, this method comprises:
A, near-end cache module are preserved near end input signal, and simultaneously, the near-end speech detection module is preserved near end input signal and judged whether current this signal is voice signal, if voice signal execution in step B; The far-end speech detection module is preserved remote end input signal and is judged whether current this signal is voice signal simultaneously, if voice signal execution in step C;
B, near-end pitch period module receive near end input signal and generate the near-end speech pitch period and send to signal separation module according near end input signal;
C, far-end pitch period module receive remote end input signal and generate the far-end speech pitch period and send to signal separation module according to remote end input signal;
D, signal separation module are carried out independent component analysis according to the near end input signal of preserving, and when the independent component analysis result restrains, calculate each signal source and pitch period thereof; And according to the near-end speech pitch period and the far-end speech pitch period that receive, the pitch period of described each signal source is compared with near-end speech pitch period and far-end speech pitch period, obtain echo signal, the near end signal that calculates after echo is eliminated exports far-end to.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910105667 CN101820302B (en) | 2009-02-27 | 2009-02-27 | Device and method for canceling echo |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910105667 CN101820302B (en) | 2009-02-27 | 2009-02-27 | Device and method for canceling echo |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101820302A true CN101820302A (en) | 2010-09-01 |
| CN101820302B CN101820302B (en) | 2013-10-30 |
Family
ID=42655265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910105667 Active CN101820302B (en) | 2009-02-27 | 2009-02-27 | Device and method for canceling echo |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101820302B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103327201A (en) * | 2012-03-20 | 2013-09-25 | 联芯科技有限公司 | Elimination method and system of remaining echoes |
| CN103347138A (en) * | 2013-06-21 | 2013-10-09 | 苏州鼎富软件科技有限公司 | Cyber dialogue echo eliminating method |
| CN107564539A (en) * | 2017-08-29 | 2018-01-09 | 苏州奇梦者网络科技有限公司 | Towards the acoustic echo removing method and device of microphone array |
| CN104410762B (en) * | 2014-11-18 | 2018-04-27 | 沈阳工业大学 | Sane echo cancelltion method in hands-free speaking system |
| CN108353007A (en) * | 2015-11-26 | 2018-07-31 | 罗伯特·博世有限公司 | Method and apparatus for signal Analysis data |
| CN109413522A (en) * | 2018-11-08 | 2019-03-01 | 深圳市云威物联科技有限公司 | Audio signal processing method, audio-signal processing apparatus and electronic equipment |
| CN110246516A (en) * | 2019-07-25 | 2019-09-17 | 福建师范大学福清分校 | The processing method of small space echo signal in a kind of voice communication |
| CN111968660A (en) * | 2019-05-20 | 2020-11-20 | 北京地平线机器人技术研发有限公司 | Echo cancellation device and method, electronic device, and storage medium |
| CN113707166A (en) * | 2021-04-07 | 2021-11-26 | 腾讯科技（深圳）有限公司 | Voice signal processing method, apparatus, computer device and storage medium |
| CN113707166B (en) * | 2021-04-07 | 2024-06-07 | 腾讯科技（深圳）有限公司 | Voice signal processing method, device, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1780165A (en) * | 2004-11-23 | 2006-05-31 | 华为技术有限公司 | Echo eliminator and elimination |
| CN1842110A (en) * | 2005-03-28 | 2006-10-04 | 华为技术有限公司 | Echo eliminating device and method |
| CN1946105A (en) * | 2006-10-27 | 2007-04-11 | 华南理工大学 | Method and its system for eliminating stereophonic echo based on voice signal separate model |
| CN101346897A (en) * | 2005-12-22 | 2009-01-14 | 冲电气工业株式会社 | Echo canceller |
-
2009
- 2009-02-27 CN CN 200910105667 patent/CN101820302B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1780165A (en) * | 2004-11-23 | 2006-05-31 | 华为技术有限公司 | Echo eliminator and elimination |
| CN1842110A (en) * | 2005-03-28 | 2006-10-04 | 华为技术有限公司 | Echo eliminating device and method |
| CN101346897A (en) * | 2005-12-22 | 2009-01-14 | 冲电气工业株式会社 | Echo canceller |
| CN1946105A (en) * | 2006-10-27 | 2007-04-11 | 华南理工大学 | Method and its system for eliminating stereophonic echo based on voice signal separate model |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103327201B (en) * | 2012-03-20 | 2016-04-20 | 联芯科技有限公司 | Residual echo removing method and system |
| CN103327201A (en) * | 2012-03-20 | 2013-09-25 | 联芯科技有限公司 | Elimination method and system of remaining echoes |
| CN103347138A (en) * | 2013-06-21 | 2013-10-09 | 苏州鼎富软件科技有限公司 | Cyber dialogue echo eliminating method |
| CN104410762B (en) * | 2014-11-18 | 2018-04-27 | 沈阳工业大学 | Sane echo cancelltion method in hands-free speaking system |
| CN108353007A (en) * | 2015-11-26 | 2018-07-31 | 罗伯特·博世有限公司 | Method and apparatus for signal Analysis data |
| CN108353007B (en) * | 2015-11-26 | 2022-01-11 | 罗伯特·博世有限公司 | Method and apparatus for analyzing signal data |
| CN107564539B (en) * | 2017-08-29 | 2021-12-28 | 苏州奇梦者网络科技有限公司 | Acoustic echo cancellation method and device facing microphone array |
| CN107564539A (en) * | 2017-08-29 | 2018-01-09 | 苏州奇梦者网络科技有限公司 | Towards the acoustic echo removing method and device of microphone array |
| CN109413522A (en) * | 2018-11-08 | 2019-03-01 | 深圳市云威物联科技有限公司 | Audio signal processing method, audio-signal processing apparatus and electronic equipment |
| CN109413522B (en) * | 2018-11-08 | 2020-09-04 | 深圳市云威物联科技有限公司 | Sound signal processing method, sound signal processing device and electronic equipment |
| CN111968660A (en) * | 2019-05-20 | 2020-11-20 | 北京地平线机器人技术研发有限公司 | Echo cancellation device and method, electronic device, and storage medium |
| CN110246516A (en) * | 2019-07-25 | 2019-09-17 | 福建师范大学福清分校 | The processing method of small space echo signal in a kind of voice communication |
| CN113707166A (en) * | 2021-04-07 | 2021-11-26 | 腾讯科技（深圳）有限公司 | Voice signal processing method, apparatus, computer device and storage medium |
| CN113707166B (en) * | 2021-04-07 | 2024-06-07 | 腾讯科技（深圳）有限公司 | Voice signal processing method, device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101820302B (en) | 2013-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101820302B (en) | Device and method for canceling echo | |
| Gustafsson et al. | Combined acoustic echo control and noise reduction for hands-free telephony | |
| Gustafsson et al. | A psychoacoustic approach to combined acoustic echo cancellation and noise reduction | |
| CN109716743B (en) | Full duplex voice communication system and method | |
| Valin et al. | Low-complexity, real-time joint neural echo control and speech enhancement based on percepnet | |
| KR100989266B1 (en) | Double talk detection method based on spectral acoustic properties | |
| US20040264610A1 (en) | Interference cancelling method and system for multisensor antenna | |
| EP0280719B1 (en) | Linear predictive echo canceller integrated with relp vocoder | |
| Park et al. | Integrated echo and noise canceler for hands-free applications | |
| JP2003500936A (en) | Improving near-end audio signals in echo suppression systems | |
| Chen et al. | Nonlinear residual echo suppression based on multi-stream conv-tasnet | |
| Ma et al. | Echofilter: End-to-end neural network for acoustic echo cancellation | |
| Lee et al. | A statistical model-based residual echo suppression | |
| CN101958122A (en) | Method and device for eliminating echo | |
| KR100386488B1 (en) | Arrangement for communication with a subscriber | |
| JP2003309493A (en) | Method, device and program for reducing echo | |
| Dreiseitel et al. | Acoustic echo and noise control—a long lasting challenge | |
| Chhetri et al. | Regression-based residual acoustic echo suppression | |
| KR100272131B1 (en) | Adaptive reverbation cancelling apparatus | |
| Jung et al. | A new adaptive algorithm for stereophonic acoustic echo canceller | |
| CN111883155A (en) | Echo cancellation method, device and storage medium | |
| Zheng et al. | Real-time speech enhancement with dynamic attention span | |
| Zahradnik et al. | A simple echo attenuation in signals | |
| Parikh et al. | Study of echo cancelling algorithms for full duplex telephone networks with vocoders | |
| Bekrani et al. | An efficient quasi LMS/newton adaptive algorithm for stereophonic acoustic echo cancellation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |