CN114464202A - Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition - Google Patents
Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition Download PDFInfo
- Publication number
- CN114464202A CN114464202A CN202210028192.1A CN202210028192A CN114464202A CN 114464202 A CN114464202 A CN 114464202A CN 202210028192 A CN202210028192 A CN 202210028192A CN 114464202 A CN114464202 A CN 114464202A
- Authority
- CN
- China
- Prior art keywords
- length
- filter
- signal
- error signal
- current
- 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.)
- Pending
Links
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000003044 adaptive effect Effects 0.000 claims abstract description 29
- 239000013598 vector Substances 0.000 claims abstract description 25
- 238000005070 sampling Methods 0.000 claims abstract description 9
- 230000017105 transposition Effects 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 2
- 238000004891 communication Methods 0.000 abstract description 9
- 230000000694 effects Effects 0.000 abstract 1
- 238000004088 simulation Methods 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000002592 echocardiography Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L2021/02082—Noise filtering the noise being echo, reverberation of the speech
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Filters That Use Time-Delay Elements (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention discloses a hyperbolic secant echo cancellation method based on nearest kronecker product decomposition, which comprises the following steps: sampling a voice signal transmitted from a far end so as to obtain a far end signal discrete value at the current n moment; sampling echo signals collected by a near-end microphone to obtain an expected signal at the current moment n; calculating the output of the whole adaptive filter; subtracting the output signal from the near-end signal to obtain an error signal, and calculating the length L1And L2The error signal of the decomposition filter of (1); calculating the length L according to the error signal of the current n time1And L2The decomposition filter at the current n moment based on the de-impact interference error signal of the nearest kronecker product arctangent function to update and obtain the de-impact interference error signal with the length L1And L2The tap weight vector of the filter at the next moment n +1 is decomposed, and the weight vector of the whole self-adaptive filter is updated; updating the value of n, and repeating the steps until the call is madeAnd (6) ending. The method has strong identification capability on telephone communication, high convergence speed, low steady-state error and obvious echo cancellation effect.
Description
Technical Field
The invention belongs to the technical field of adaptive echo cancellation of voice communication, and particularly relates to a generalized hyperbolic secant adaptive echo cancellation method based on nearest kronecker product decomposition.
Background
In a communication system, noise and echo interference cannot always be ignored. When using wired, wireless, network and other communication devices, users occasionally hear their own voice at the receiving end, which is called echo phenomenon and is also the largest interference affecting the call quality. For example, acoustic echo is often generated when a multi-person network audio conference is held or a user uses a hands-free function of a communication device. The principle of the generation is that the voice signal of the caller is picked up by a microphone, transmitted to the near end, amplified by a loudspeaker and output. An echo is generated in the near-end room, and the echo signal is picked up by the near-end microphone and transmitted back to the far-end output, so that the speaker hears his own voice. The short echo delay can be hardly perceived, and can be understood as a form of spectral distortion. Conversely, the echo is clearly noticeable with a delay of more than a few tens of milliseconds. Under extreme conditions, when the echo signal gain is too large to form positive feedback, harsh howling will be caused, and communication cannot be performed. Therefore, an Acoustic Echo Cancellation (AEC) must be integrated into the communication device to suppress the Echo and improve the communication quality. Since the human ear is extremely sensitive to echoes, the study of methods for eliminating acoustic echoes is still a popular topic. One of typical widely used adaptive echo cancellation methods is the Least-Mean-Square (LMS) algorithm, but the conventional algorithm becomes very unstable when there is interference noise such as impulse noise in the system.
The current mature echo cancellation methods against interference are as follows: document 1, "near Kronecker product composition based normalized least squares mean algorithm" (Bhattacharjee, Sankha Subhra, and Nithin V.George., IEEE International Conference on Acoustics, speed and Signal Processing (ICASSP), pp.476-480.IEEE,2020.) this method utilizes normalized least mean square criteria to achieve fast convergence. But since the algorithm does not take into account the impulse noise characteristics in the system, the performance will be degraded or even spread out when solving the problem of systems containing impulse noise.
Disclosure of Invention
Aiming at the defects or improvement requirements of the prior art, the invention provides a generalized hyperbolic secant self-adaptive echo cancellation method based on the nearest kronecker product decomposition.
In order to achieve the above object, the present invention provides a generalized hyperbolic secant adaptive echo cancellation method based on nearest kronecker product decomposition, comprising:
s1: sampling a voice signal transmitted from a far end so as to obtain a far end signal discrete value x (n) at the current n moment; meanwhile, the echo signal collected by the near-end microphone is sampled to obtain the expected signal at the current n moment
S2: a passage length of L1And L2The decomposition filter of (2) calculates the output y (n) of the entire adaptive filter;
s3: the near-end signalSubtracting the output signal y (n) to obtain an error signal e (n), and calculating the length L1And L2Of the decomposition filter e1(n) and e2(n);
S4: calculating the length L according to the error signal e (n) at the current time n1And L2The decomposition filter based on the nearest kronecker product arctangent function at the current n timeAndupdating by using de-impact interference error signal to obtain length L1And L2The tap weight vector W at the next time instant n +1 of the decomposition filter1(n +1) and W2(n +1), further realizing the update of the weight vector of the whole adaptive filter;
s5: updating n to the next time value, and repeatedly executing the steps S1-S4 until the call is ended.
In some alternative embodiments, step S2 includes:
s2.1: the discrete value of the far-end signal from the current n moment to the n-L +1 moment forms the whole self of the current n momentAn adaptive filter input vector x (n), x (n) ═ x (n), x (n-1)]TWherein, L represents the tap number of the adaptive filter, and the superscript T represents transposition;
s2.2: a passage length of L1And L2The decomposition filter of (2) calculates the output y (n) of the entire adaptive filter at the current time instant n.
In some alternative embodiments, y (n) ═ WT(n) x (n), wherein,the tap weight vector of the whole adaptive filter at the current n time is zero and has the length equal to L ═ L1×L2,W2,d(n) and W1,d(n) each represents a length L1And L2Decomposing the impulse response of the filter;representing the kronecker product operation and D the number of decomposition filters.
In some alternative embodiments, step S3 includes:
the expected signal of the current n timeSubtracting the output signal y (n) of the whole adaptive filter at the current time n to obtain an error signal e (n) at the current time n, and transmitting the error signal e (n) to the far end as the near end signal with echo removed at the current time n, wherein,
determining the length L from the error signal e (n) at the current time n1And L2The error signal of the decomposition filter.
In some alternative embodiments, the composition is prepared byTo a length L1By decomposing the error signal of the filterTo a length L2The error signal of the decomposition filter of (1), wherein W1(n)=[W1,1(n),...,W1,d(n),...,W1,D(n)]T,W2(n)=[W2,1(n),...,W2,d(n),...,W2,D(n)]TRespectively represent a length L2And L1Decomposing the weight vector of the filter; corresponding length L2And L1The inputs to the decomposition filters are respectively:parameter(s)And let D be L2, Respectively represent a length L2And L1The unit vector of (2).
In some alternative embodiments, step S4 includes:
s4.1: according to the error signal e (n) of the current n time, the error signal for removing the impact interference at the current n time is calculatedIn which, among others,is expressed as length L1The de-impact of the decomposition filter of (1) the interfering error signal,is expressed as length L2The decomposition filter of (1) represents an error signal for removing the impulse interference, sech (-) represents a hyperbolic secant operation, and tanh (-) represents a hyperbolic secant operationPerforming curve tangent operation;the initial values are zero, and lambda and alpha are parameters for controlling the shape of the hyperbolic secant function;
s4.2: updated to respectively obtain a length L2And L1The tap weight vector W at the next time instant n +1 of the decomposition filter1(n +1) and W2(n +1), thereby implementing a weight vector update for the entire adaptive filter, wherein,μ is the step size parameter.
In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects:
the value of the generalized hyperbolic secant function of the present invention varies according to the state of the noise environment, and when there is no impulse noise,when there is an impact noise present, the noise,close to zero. In other words, when impact noise exists, the algorithm is not updated, which shows that the algorithm has good impact noise resistance capability, and a smaller steady-state error can be obtained; when there is no impact noise, the formula is updatedThe term is close to e (n). Therefore, the algorithm can obtain a faster convergence speed and has good impact-resistant robust performance. In conclusion, the method provided by the invention can adjust the contradiction between the high convergence rate and the low steady-state error and resist impact noise.
Drawings
Fig. 1 is a flowchart of a generalized hyperbolic secant adaptive echo cancellation method based on nearest kronecker product decomposition according to an embodiment of the present invention;
FIG. 2 is a channel diagram of a simulation experiment provided by an embodiment of the present invention;
fig. 3 is a normalized steady-state imbalance curve of a simulation experiment in a NKP-NLMS method and a method according to the present invention when a real speech signal is an input signal according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
As shown in fig. 1, a hyperbolic secant echo cancellation method based on nearest kronecker product decomposition according to an embodiment of the present invention includes the following steps:
s1: sampling signal
Sampling a voice signal transmitted from a far end to obtain a far end signal discrete value x (n) at the current n moment; meanwhile, the echo signal collected by the near-end microphone is sampled to obtain the expected signal at the current n moment
S2: calculating the output y (n) of the whole adaptive filter;
s2.1: the discrete values of the far-end signals from the current time n to the time n-L +1 form the whole input vector x (n) of the adaptive filter at the current time n, wherein x (n) is [ x (n), x (n-1)..., x (n-L +1)]TWherein, L represents the number of taps of the adaptive filter, L is 50 or 500, and the superscript T represents transposition;
s2.2: a passage length of L1And L2The decomposition filter of (2) calculates the output y (n) of the entire adaptive filter at the current time n, i.e., y (n) ═ WT(n) x (n), wherein,the tap weight vector of the whole adaptive filter at the current n time is zero and has the length equal to L ═ L1×L2;W2,d(n) and W1,d(n) each represents a length L1And L2Decomposing the impulse response of the filter;representing the kronecker product operation and D the number of decomposition filters. (ii) a
S3: echo cancellation
The expected signal of the current n timeSubtracting the output signal y (n) of the whole adaptive filter at the current n moment to obtain an error signal e (n) at the current n moment, and transmitting the error signal e (n) serving as a near-end signal with echo removed at the current n moment to a far end, namelyThen the length is L1And L2The error signal of the decomposition filter of (a) can be written in two equal forms, namely:
wherein, W1(n)=[W1,1(n),...,W1,d(n),...,W1,D(n)]T,
W2(n)=[W2,1(n),...,W2,d(n),...,W2,D(n)]TRespectively represent a length L2And L1Decomposing the weight vector of the filter; corresponding length L2And L1The inputs to the decomposition filters are respectively:
s4: updating of weight vectors
S4.1: according to the error signal e (n) of the current n time, the error signal for removing the impact interference at the current n time is calculatedOf two equal forms, i.e. of length L1And L2The decomposition filter of
Wherein sech (·) represents hyperbolic secant operation; tanh (·) represents a hyperbolic tangent operation;the initial values are zero, and lambda and alpha are parameters for controlling the shape of the hyperbolic secant function;
s4.2: updated to respectively obtain a length L2And L1The tap weight vector W at the next time instant n +1 of the decomposition filter1(n +1) and W2(n +1), furtherUpdating the weight vector of the whole adaptive filter;
wherein mu is a step length parameter and takes a value of 0.5;
s5: and repeating the steps S1 to S4 until the call is ended, with n being equal to n + 1.
Simulation experiment
In order to verify the effectiveness of the present invention, a simulation experiment was performed and compared with the method of prior document 1.
The echo channel impulse response of the simulation experiment is obtained in a quiet closed room with the length of 6.25m, the width of 3.75m, the height of 2.5m, the temperature of 20 ℃ and the humidity of 50 percent, and the impulse response length, namely the number L of taps of the filter is 50. The background noise is zero-mean white Gaussian noise with a signal-to-noise ratio of 30dB and the sampling frequency is 8 KHz.
According to the above experimental conditions, the echo cancellation experiment was performed by the method of the present invention and the method of prior document 1. The experimental optimal parameter values for each method are shown in table 1.
TABLE 1 values of the optimum parameters for the experiments of the methods
Document 1(NKP-NLMS) | μ=0.15,L1=10,L2=5 |
The invention | μ=0.5,α=0.01,λ=0.2,L1=10,L2=5 |
Fig. 2 is a channel diagram of a communication system constituted by a quiet closed room for experiment.
Fig. 3 shows a normalized steady-state imbalance curve obtained by a simulation experiment when a real speech signal is an input signal according to the method of document 1(NKP-NLMS) and the method of the present invention.
As can be seen from fig. 3: the invention converges at about 1000 sampling moments, and the steady state error is about-30 dB; while document 1 is directly divergent at about 480 sampling instants; the performance of the invention is clearly optimal in systems containing impulsive noise.
It should be noted that, according to the implementation requirement, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can be combined into new steps/components to achieve the purpose of the present invention.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (6)
1. A hyperbolic secant echo cancellation method based on nearest-Kernel product decomposition is characterized by comprising the following steps:
s1: sampling a voice signal transmitted from a far end so as to obtain a far end signal discrete value x (n) at the current n moment; meanwhile, the echo signal collected by the near-end microphone is sampled to obtain the expected signal at the current n moment
S2: a passage length of L1And L2The decomposition filter of (2) calculates the output y (n) of the entire adaptive filter;
s3: the near-end signalSubtracting the output signal y (n) to obtain an error signal e (n), and calculating the length L1And L2Of the decomposition filter e1(n) and e2(n);
S4: calculating the length L according to the error signal e (n) at the current time n1And L2The decomposition filter based on the nearest kronecker product arctangent function at the current n timeAndupdating by using de-impact interference error signal to obtain length L1And L2The tap weight vector W at the next time instant n +1 of the decomposition filter1(n +1) and W2(n +1), further realizing the update of the weight vector of the whole adaptive filter;
s5: updating n to the next time value, and repeatedly executing the steps S1-S4 until the call is ended.
2. The method according to claim 1, wherein step S2 includes:
s2.1: the discrete values of the far-end signals from the current time n to the time n-L +1 form the whole input vector x (n) of the adaptive filter at the current time n, wherein x (n) is [ x (n), x (n-1)..., x (n-L +1)]TWherein, L represents the tap number of the adaptive filter, and superscript T represents transposition;
s2.2: a passage length of L1And L2The decomposition filter of (2) calculates the output y (n) of the entire adaptive filter at the current time instant n.
3. The method of claim 2, step S2.2 comprising:
y(n)=WT(n) x (n), wherein,the tap weight vector of the whole adaptive filter at the current n time is zero and has the length equal to L ═ L1×L2,W2,d(n) and W1,d(n) each represents a length L1And L2Decomposing the impulse response of the filter;representing the kronecker product operation and D the number of decomposition filters.
4. The method according to claim 3, wherein step S3 includes:
the expected signal of the current n timeSubtracting the output signal y (n) of the whole adaptive filter at the current time n to obtain an error signal e (n) at the current time n, and transmitting the error signal e (n) to the far end as the near end signal with echo removed at the current time n, wherein,
determining the length L from the error signal e (n) at the current time n1And L2The error signal of the decomposition filter.
5. The method of claim 4, wherein the method is performed byTo a length L1By decomposing the error signal of the filterTo a length L2The error signal of the decomposition filter of (1), wherein W1(n)=[W1,1(n),...,W1,d(n),...,W1,D(n)]T,W2(n)=[W2,1(n),...,W2,d(n),...,W2,D(n)]TRespectively represent a length L2And L1Decomposing the weight vector of the filter; corresponding length L2And L1The inputs to the decomposition filters are respectively:parameter(s)And let D be L2,Respectively represent a length L2And L1The unit vector of (2).
6. The method according to claim 5, wherein step S4 includes:
s4.1: according to the error signal e (n) of the current n time, the error signal for removing the impact interference at the current n time is calculatedIn which, among others,is expressed as length L1The de-impact of the decomposition filter of (1) the interfering error signal,is expressed as length L2The error signal of the decomposition filter for removing the impact interference of (1), sech (·) represents hyperbolic secant operation, tanh (·) represents hyperbolic tangent operation;the initial values are zero, and lambda and alpha are parameters for controlling the shape of the hyperbolic secant function;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210028192.1A CN114464202A (en) | 2022-01-11 | 2022-01-11 | Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210028192.1A CN114464202A (en) | 2022-01-11 | 2022-01-11 | Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114464202A true CN114464202A (en) | 2022-05-10 |
Family
ID=81409855
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210028192.1A Pending CN114464202A (en) | 2022-01-11 | 2022-01-11 | Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114464202A (en) |
-
2022
- 2022-01-11 CN CN202210028192.1A patent/CN114464202A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110838300B (en) | Echo cancellation processing method and processing system | |
US6961422B2 (en) | Gain control method for acoustic echo cancellation and suppression | |
US11297178B2 (en) | Method, apparatus, and computer-readable media utilizing residual echo estimate information to derive secondary echo reduction parameters | |
EP2045929B1 (en) | Multi-channel echo cancellation with round robin regularization | |
US20060018457A1 (en) | Voice activity detectors and methods | |
CN106448691B (en) | Voice enhancement method for public address communication system | |
EP2101480A2 (en) | Echo canceller and echo cancelling method | |
CN112689056B (en) | Echo cancellation method and echo cancellation device using same | |
US9020144B1 (en) | Cross-domain processing for noise and echo suppression | |
CN110956975A (en) | Echo cancellation method and device | |
CN109697986B (en) | Adaptive bias compensation echo cancellation method based on minimum cubic absolute value | |
CN107871510A (en) | A kind of zero attracts to become the maximum entropy echo cancel method of core width | |
CN115985278A (en) | Improved convex combination decorrelation proportional adaptive echo cancellation method | |
CN113409806B (en) | Zero-attraction echo cancellation method based on arctangent function | |
CN114464202A (en) | Hyperbolic secant echo cancellation method based on nearest kronecker product decomposition | |
Raghavendran | Implementation of an acoustic echo canceller using matlab | |
CN110767245B (en) | Voice communication self-adaptive echo cancellation method based on S-shaped function | |
Iwai et al. | Acoustic echo and noise canceller using shared-error normalized least mean square algorithm | |
Fukui et al. | Acoustic echo canceller software for VoIP hands-free application on smartphone and tablet devices | |
Kotte | Performance Analysis of Adaptive Algorithms based on different parameters Implemented for Acoustic Echo Cancellation in Speech Signals | |
Motar et al. | Echo Cancellation in Telecommunications Using Variable Step-Size, Dynamic Selection, Affine Projection Algorithm. | |
Gunale et al. | Frequency domain adaptive filter using FFT algorithm for acoustic echo cancellation | |
CN117238305A (en) | Point-amplitude sub-band block proportional self-adaptive acoustic echo cancellation method | |
Szabolcs et al. | Hands-Free VoIP Terminal with Gain Control Based on Neural Network | |
Zhu et al. | A Partitioned-Block Frequency-Domain Adaptive Kalman Filter for Stereophonic Acoustic Echo Cancellation. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20231116 Address after: Hengzhang village, Shiqi street, Haishu District, Ningbo City, Zhejiang Province Applicant after: Ningbo Lidou Intelligent Technology Co.,Ltd. Applicant after: WUHAN University Address before: Hengzhang village, Shiqi street, Haishu District, Ningbo City, Zhejiang Province Applicant before: Ningbo Lidou Intelligent Technology Co.,Ltd. |
|
TA01 | Transfer of patent application right |