A data interpolation echo cancellation method and system for telephone communication

Abstract

The present application relates to the field of telephone communication, and specifically provides a data interpolation echo cancellation method and system for telephone communication, the method comprising: obtaining an actual random missing input signal and an expected signal; obtaining an input signal missing interpolation value according to the actual random missing input signal and an interpolation model; constructing an input missing interpolation vector according to the input signal missing interpolation value; generating an interpolation model error signal according to the input missing interpolation vector and the expected signal; generating a variance estimate of the error signal variance according to the interpolation model error signal; generating an optimal step length according to the variance estimate; smoothing the optimal step length to obtain a step length value; obtaining an unbiased estimate of the actual gradient according to the missing data interpolation method, the input missing interpolation vector and the variance estimate; and updating the weight vector of the echo canceller according to the unbiased estimate and the step length value. The method can simultaneously have a faster convergence speed and a lower steady-state misalignment, and has a stronger anti-pulse performance.

Description

Technical Field

[0001] The present invention relates to the field of telephone communication, and in particular to a data interpolation echo elimination method and system for telephone communication. Background Art

[0002] In telephone communication, when one party (Party A) speaks, the voice signal is collected by the microphone and transmitted to the speaker of the other party (Party B). If the microphone of Party B also captures this voice signal and transmits it back to Party A, an echo will be formed at Party A. Echo cancellation technology is used to improve the call quality of telephone communication devices such as hands-free phones and mobile phones. By continuously adjusting the weight vector of the echo canceller, the technology can automatically track and estimate the impulse response of the echo path, generate an approximate signal of the echo signal, and finally eliminate the echo from the voice signal of the call.

[0003] In the process of echo cancellation in telephone communication, random missing of input voice signals is a common problem. This missing may be caused by channel interference, data packet loss or equipment failure, which will have a significant impact on the performance of the echo canceller. The missing signal will cause the echo canceller to be unable to accurately estimate the impulse response of the echo path, thereby reducing its performance, leaving a noticeable echo in the call, affecting the user experience. In addition, frequent data loss will reduce the stability and reliability of the entire telephone communication system. In order to meet these challenges, effective data interpolation technology must be used to fill the missing voice signal to ensure the normal operation of the echo canceller and improve the call quality.

[0004] In addition, in some special environments, the output signal of the unknown system may be contaminated by impulse noise. Impulse noise is a transient high-energy interference signal caused by factors such as electromagnetic interference, switching power supply noise or equipment failure. This interference will prevent the echo canceller from accurately estimating the impulse response of the echo path, reducing its effectiveness, resulting in obvious echo and noise in the call, affecting the user experience. In addition, frequent impulse noise interference will increase the computational burden of the system and reduce the stability and reliability of the telephone communication system. Therefore, effective noise detection and suppression and robust algorithm design are key measures to mitigate the impact of impulse noise.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] To this end, the technical problem to be solved by the present invention is to overcome the problem in the prior art that the output signal of an unknown system may be contaminated by pulse noise in some special environments. Pulse noise is a transient high-energy interference signal caused by factors such as electromagnetic interference, switching power supply noise or equipment failure. This interference will prevent the echo canceller from accurately estimating the pulse response of the echo path, reducing its effectiveness, resulting in obvious echo and noise during calls, affecting the user experience. In addition, frequent pulse noise interference will increase the computational burden of the system and reduce the stability and reliability of the telephone communication system.

[0007] In order to solve the above technical problems, the first aspect of the present invention provides a data interpolation echo cancellation method for telephone communication, the method comprising:

[0008] Get the actual random missing input signal and the expected signal;

[0009] Obtaining an input signal missing interpolation value according to the actual randomly missing input signal and the interpolation model;

[0010] Constructing an input missing interpolation vector according to the input signal missing interpolation value;

[0011] generating an interpolation model error signal according to the input missing interpolation vector and the expected signal;

[0012] generating a variance estimate of an error signal variance based on the interpolation model error signal;

[0013] generating an optimal step size according to the variance estimate;

[0014] Smoothing the optimal step length to obtain a step length value;

[0015] Obtaining an unbiased estimate of the actual gradient based on the missing data interpolation method, the input missing interpolation vector, and the variance estimate;

[0016] The weight vector of the echo canceller is updated according to the unbiased estimation value and the step value.

[0017] In one embodiment of the present invention, the input missing interpolation vector is expressed as:

[0018] ;

[0019] in, for Time and its previous continuous The input signal at the moment is missing the interpolation value, Represents the weight vector length of the echo canceller, the superscript Represents a transpose operation.

[0020] In one embodiment of the present invention, the interpolation model error signal is expressed as:

[0021] ;

[0022] in, for The expected signal of the moment, for Time Echo Canceller Tap coefficients The weight vector formed by is the probability that the input data is missing randomly, , is a value between A positive constant in the middle.

[0023] In one embodiment of the present invention, the unbiased estimate of the actual gradient is expressed as:

[0024] ;

[0025] in, is an estimate of the variance of the error signal, represents the diagonal matrix obtained by arranging the vectors on the diagonal, .

[0026] In one embodiment of the present invention, the variance estimate is expressed as:

[0027] ;

[0028] in, For values ​​between Intermediate smoothing factor.

[0029] In one embodiment of the present invention, the optimal step length is expressed as:

[0030] ;

[0031] in, is the noise variance, , is the first element of the autocorrelation matrix of the input signal, .

[0032] In one embodiment of the present invention, the step value is expressed as:

[0033] ;

[0034] in, For values ​​between The smoothing factor in the middle, is the upper limit of the variable step value, It means taking the minimum value of the items in the brackets.

[0035] In one embodiment of the present invention, the formula for updating the weight vector of the echo canceller is:

[0036] ;

[0037] in, for The step value at time, is an unbiased estimate of the actual gradient.

[0038] A second aspect of the present invention provides a data interpolation echo cancellation system for telephone communication, the system comprising: a vector construction module, an estimation value calculation module, a step length calculation module, and an update module;

[0039] The vector construction module is configured to: obtain an actual random missing input signal and an expected signal; obtain an input signal missing interpolation value according to the actual random missing input signal and an interpolation model; and construct an input missing interpolation vector according to the input signal missing interpolation value;

[0040] The estimated value calculation module is configured to: generate an interpolation model error signal according to the input missing interpolation vector and the expected signal; generate a variance estimate of the error signal variance according to the interpolation model error signal;

[0041] The step length calculation module is configured to: generate an optimal step length according to the variance estimation value; smooth the optimal step length to obtain a step length value;

[0042] The updating module is configured to: obtain an unbiased estimate of the actual gradient according to the missing data interpolation method, the input missing interpolation vector and the variance estimate; and update the weight vector of the echo canceller according to the unbiased estimate and the step value.

[0043] The above technical solution of the present invention has the following advantages compared with the prior art:

[0044] The data interpolation echo cancellation method and system for telephone communication described in the present invention not only has a strong ability to resist impulse noise, but also can accelerate the convergence speed of the echo canceller when the input voice signal is randomly missing. It can simultaneously have a faster convergence speed and a lower steady-state imbalance, and has a strong anti-impulse performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.

[0046] Figure 1 It is a flow chart of a data interpolation echo cancellation method for telephone communication provided by the present invention;

[0047] Figure 2 It is a comparative schematic diagram of a data interpolation echo cancellation method for telephone communication and a normalized mean square deviation curve of an echo cancellation system in the system provided by the present invention;

[0048] Figure 3 It is an architecture diagram of a data interpolation echo cancellation system for telephone communication provided by the present invention. DETAILED DESCRIPTION

[0049] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0050] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0051] Reference Figure 1 As shown, the present invention provides a data interpolation echo cancellation method for telephone communication, the method comprising:

[0052] S100, obtaining an actual random missing input signal and an expected signal;

[0053] In step S100, the actual random missing input signal is obtained and expected signal .

[0054] S200, obtaining an input signal missing interpolation value according to the actual randomly missing input signal and the interpolation model;

[0055] In step S200, an interpolation model and actual random missing input signals are used to generate missing interpolation values ​​of input signals, and the interpolation model is: ,or ,in, is the actual random missing input signal, is an independent and identically distributed Bernoulli random variable, that is, The probabilities of taking 1 and 0 are and , is the original input signal without any loss.

[0056] S300, constructing an input missing interpolation vector according to the input signal missing interpolation value;

[0057] In step S300, the input missing interpolation vector is expressed as:

[0058] (1);

[0059] in, for Time and its previous continuous The input signal at the moment is missing the interpolation value, Represents the weight vector length of the echo canceller, the superscript Represents a transpose operation.

[0060] In actual application scenarios, Time and its previous continuous The input signal at the moment is missing interpolation value Construct the input missing interpolation vector, where the superscript Represents a transpose operation.

[0061] S400, generating an interpolation model error signal according to the input missing interpolation vector and the expected signal;

[0062] In step S400, the error signal is expressed as:

[0063] (2);

[0064] in, for Time Echo Canceller Tap coefficients The weight vector formed.

[0065] In actual application scenarios, the missing interpolation vector is Construct two vectors related to it respectively as well as ,in, is the probability that the input data is missing randomly, For values ​​in the interval A positive constant within . Input missing interpolation vector and expected signal at each moment Compute the error signal of the interpolation model, where for Time Echo Canceller Tap coefficients The weight vector formed.

[0066] S500, generating a variance estimate of an error signal variance according to the interpolation model error signal;

[0067] In step S500, the variance estimate is expressed as:

[0068] (3);

[0069] in, For values ​​between Intermediate smoothing factor.

[0070] In actual application scenarios, the variance estimate of the error signal variance is obtained based on the interpolation model error signal and formula (3): .in, For values ​​between Intermediate smoothing factor.

[0071] S600, generating an optimal step length according to the variance estimation value;

[0072] In step S600, the optimal step length is expressed as:

[0073] (4);

[0074] in, is the noise variance, , is the first element of the autocorrelation matrix of the input signal, .

[0075] In actual application scenarios, the optimal step size of the echo canceller after missing data interpolation is obtained based on the variance estimate and formula (4).

[0076] S700, smoothing the optimal step length to obtain a step length value;

[0077] In step S700, the step value is expressed as:

[0078] (5);

[0079] in, For values ​​between The smoothing factor in the middle, is the upper limit of the variable step value, It means taking the minimum value of the items in the brackets.

[0080] In actual application scenarios, the optimal step length is smoothed using formula (5) to obtain the step length value ,in, For values ​​between The smoothing factor in the middle, is the upper limit of the variable step value, It means taking the minimum value of the items in the brackets.

[0081] S800, obtaining an unbiased estimate of the actual gradient according to the missing data interpolation method, the input missing interpolation vector and the variance estimate;

[0082] In step S800, the unbiased estimate is expressed as:

[0083] (6);

[0084] in, is an estimate of the variance of the error signal, Represents a diagonal matrix obtained by arranging the vectors on the diagonal.

[0085] In practical application scenarios, missing data interpolation is used to correct the gradient descent step to obtain an unbiased estimate of the actual gradient , the calculation formula is shown in formula (6), where: Represents a diagonal matrix obtained by arranging the vectors on the diagonal.

[0086] S900: Update a weight vector of an echo canceller according to the unbiased estimate and the step value.

[0087] In step S900, the formula for updating the weight vector of the echo canceller is:

[0088] (7);

[0089] in, for The step value at time, is an unbiased estimate of the actual gradient.

[0090] In actual application scenarios, according to the stochastic gradient descent method, the weight vector of the echo canceller is updated using formula (7), where: for The step size at each moment. When the echo canceller senses missing data, it makes full use of the correlation of the signal to interpolate the missing input signal and improve the convergence speed of the echo canceller. Speech signals in a short period of time often show periodicity. This signal characteristic helps to restore the original signal as much as possible when data is missing. The echo canceller only uses available data and the Bernoulli model of random missing data in the gradient descent step to find an unbiased estimate or a nearly unbiased estimate of the actual gradient, thereby reducing or avoiding the impact of missing input data on performance. The step size value of updating the filter coefficients is adjusted in real time by a model-driven method to further improve the performance of the echo canceller.

[0091] In order to prove the effectiveness of the present invention, this embodiment uses a computer experiment method to verify the performance of the VSS-ImdRAF echo cancellation method provided by the present invention.

[0092] The experiment estimates the unknown echo path in the echo cancellation application scenario in telephone communication under the environment of pulse noise interference and random missing input speech signal, and compares its performance with that of LMS, GMCC and ImdLMS methods.

[0093] The noise signal in this embodiment is Gaussian noise plus impulse noise, the length of the weight vector of the unknown echo path is 128, and the input signal is a speech signal with random data missing. The experiment uses the normalized mean square deviation (NMSD) as the performance measure, that is, , the unit is ,in represents taking the logarithm, is the echo path of the actual system.

[0094] The noise signal used in the experiment contains a zero mean and a variance of Gaussian white noise and an impulse noise ,Right now . Impulse noise Produced by a Bernoulli Gaussian process, that is ,in is a Bernoulli process, and the probability of taking 0 is 0.99 and the probability of taking 1 is 0.01. is a zero-mean Gaussian white noise. The simulation results of telephone communication echo cancellation refer to Figure 2 The parameters of each algorithm are LMS ( ), ImdLMS ( ), GMCC ( ), ImdRAF ( ), VSS-ImdRAF ( ). The VSS-ImdRAF echo cancellation system of the embodiment of the present application has good anti-pulse performance and has the lowest steady-state imbalance in the telephone communication application scenario where the input data is randomly missing.

[0095] Second, refer to Figure 3 As shown, the present application provides a data interpolation echo cancellation system for telephone communication, the system comprising: a vector construction module 100, an estimation value calculation module 200, a step length calculation module 300, and an update module 400;

[0096] The vector construction module 100 is configured to: obtain an actual random missing input signal and an expected signal; obtain an input signal missing interpolation value according to the actual random missing input signal and an interpolation model; and construct an input missing interpolation vector according to the input signal missing interpolation value;

[0097] The estimated value calculation module 200 is configured to: generate an interpolation model error signal according to the input missing interpolation vector and the expected signal; generate a variance estimate of the error signal variance according to the interpolation model error signal;

[0098] The step length calculation module 300 is configured to: generate an optimal step length according to the variance estimation value; smooth the optimal step length to obtain a step length value;

[0099] The updating module 400 is configured to: obtain an unbiased estimate of the actual gradient according to the missing data interpolation method, the input missing interpolation vector and the variance estimate; and update the weight vector of the echo canceller according to the unbiased estimate and the step value.

[0100] The effects of the above-mentioned system when applying the above-mentioned method can be found in the description of the above-mentioned method embodiment, which will not be repeated here.

[0101] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0102] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0105] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A data interpolation echo cancellation method for telephone communication, characterized in that: The method comprises: Get the actual random missing input signal and the expected signal; Obtaining an input signal missing interpolation value according to the actual randomly missing input signal and the interpolation model; Constructing an input missing interpolation vector according to the input signal missing interpolation value; generating an interpolation model error signal according to the input missing interpolation vector and the expected signal; generating a variance estimate of an error signal variance based on the interpolation model error signal; generating an optimal step size according to the variance estimate; Smoothing the optimal step length to obtain a step length value; Obtaining an unbiased estimate of the actual gradient based on the missing data interpolation method, the input missing interpolation vector, and the variance estimate; updating a weight vector of an echo canceller according to the unbiased estimate and the step value; The input missing interpolation vector is expressed as: ； in, for Time and its previous continuous The input signal at the moment is missing the interpolation value, Represents the weight vector length of the echo canceller, the superscript Represents the transpose operation; The interpolation model error signal is expressed as: ； in, for The expected signal of the moment, for Time Echo Canceller Tap coefficients The weight vector formed by is the probability that the input data is missing randomly, , is a value between The positive constant in the middle; The unbiased estimate of the actual gradient is expressed as: ； in, is an estimate of the variance of the error signal, represents the diagonal matrix obtained by arranging the vectors on the diagonal, ; The variance estimate is expressed as: ； in, For values ​​between The smoothing factor in the middle; The optimal step length is expressed as: ； in, is the noise variance, , is the first element of the autocorrelation matrix of the input signal, ; The step value is expressed as: ； in, For values ​​between The smoothing factor in the middle, is the upper limit of the variable step value, It means taking the minimum value of each item in brackets; The formula for updating the weight vector of the echo canceller is: ； in, for The step value at time, is an unbiased estimate of the actual gradient.

2. A data interpolation echo cancellation system for telephone communication, used to execute the data interpolation echo cancellation method for telephone communication according to claim 1, characterized in that: The system comprises: a vector construction module, an estimated value calculation module, a step length calculation module, and an update module; The vector construction module is configured to: obtain an actual random missing input signal and an expected signal; obtain an input signal missing interpolation value according to the actual random missing input signal and an interpolation model; and construct an input missing interpolation vector according to the input signal missing interpolation value; The estimated value calculation module is configured to: generate an interpolation model error signal according to the input missing interpolation vector and the expected signal; generate a variance estimate of the error signal variance according to the interpolation model error signal; The step length calculation module is configured to: generate an optimal step length according to the variance estimation value; smooth the optimal step length to obtain a step length value; The updating module is configured to: obtain an unbiased estimate of the actual gradient according to the missing data interpolation method, the input missing interpolation vector and the variance estimate; and update the weight vector of the echo canceller according to the unbiased estimate and the step value.

Images