Wide linear adaptive filter capable of resisting impulse noise interference

Abstract

The invention relates to the technical field of adaptive filters, and discloses a wide linear adaptive filter, which comprises the following steps of: performing tensor integral solution on a system weight vector at the moment of the wide linear adaptive filter to obtain two adaptive weight vectors; constructing an input matrix based on the sampling value of the input signal, and calculating an estimation signal of the moment based on the product of the input matrix of the moment and the two adaptive weight vectors; and calculating the difference value between the expected signal and the estimation signal at the moment, obtaining an estimation error signal, and updating the gain vector at the obtaining moment and the estimation error signal by using a minimum error mode criterion, updating the two adaptive weight vectors at the obtaining moment, and reconstructing the system weight vector at the moment of the wide linear adaptive filter. According to the invention, the recursive adaptive filtering algorithm resistant to impulse noise interference is realized with low calculation complexity by using weight decomposition and selection of a robust cost function.

Description

Technical Field

[0001] The present invention relates to the technical field of adaptive filters, and in particular to a wide linear adaptive filter capable of resisting impulse noise interference. Background Art

[0002] System identification is a key branch of adaptive signal processing. Many traditional problems, such as adaptive channel equalization, adaptive noise cancellation, adaptive echo cancellation, and active noise control, can be attributed to system identification. The wide linear adaptive filter is a complex system identification model characterized by its ability to simultaneously filter the complex input signal vector and its conjugate vector. This filter fully utilizes the second-order non-circular nature of complex signals, overcoming the limitation of strictly linear adaptive filters, which can only process circular complex signals. The non-circularity of complex signals can be attributed to unequal energies between the real and imaginary components or correlation between the two components.

[0003] Commonly used wide linear adaptive filters include the wide linear least mean square (WL-LMS) adaptive filter and the wide linear recursive least square (WL-RLS) adaptive filter. The former uses stochastic gradient descent to update the filter weights based only on the current input and expected signals, resulting in low computational complexity but slow convergence. The latter, on the other hand, updates the filter weights based on the input and expected signals at all known moments, resulting in rapid convergence but a high computational complexity. The WL-LMS filter's convergence speed slows further when a colored signal is input, while the WL-RLS adaptive filter remains unaffected, offering significant convergence advantages.

[0004] However, the WL-RLS adaptive filter, based on the minimum mean square error criterion, has good convergence performance in noise environments that approximately conform to a Gaussian model. However, in the presence of severe non-Gaussian noise, such as impulse noise, its performance degrades sharply, or even diverges, making it unrobust. To address this issue, several robust recursive adaptive filters have been proposed, such as the recursive maximum correlation entropy adaptive filter. These filters introduce nonlinear processing of the error signal, which can suppress the destructive effects of impulse noise. However, the computational complexity of the nonlinear function is high, which hinders many low-cost, miniaturized applications.

[0005] In the WL-RLS adaptive filter, convergence accuracy and tracking speed are in conflict. Higher accuracy requires increased memory for past information, but this reduces the tracking speed of time-varying systems and increases the adaptive filter's latency. Furthermore, when the unknown system's impulse response is long, the computational complexity of updating the inverse correlation matrix increases significantly, making it a challenging problem. Recursive least squares adaptive filters, when dealing with unknown systems with long impulse responses, suffer from high computational complexity, slow tracking speed, and poor filtering robustness.

[0006] In summary, existing wide linear adaptive filters need to process the signal vector and its conjugate vector simultaneously. The high-dimensional matrix operations in the complex domain increase the computational complexity and reduce the filtering efficiency. They cannot effectively track the updates of the system weight vector and the inverse correlation matrix in the time-varying system. In addition, impulse noise affects the system update, resulting in poor system robustness and filtering distortion. Summary of the Invention

[0007] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the existing technology cannot take into account both computational complexity and filtering robustness, resulting in slow filtering speed and poor filtering effect.

[0008] To solve the above technical problems, the present invention provides a wide linear adaptive filter resistant to impulse noise interference, comprising: Wide Linear Adaptive Filter The system weight vector at time , is decomposed into tensors to obtain The first adaptive weight vector and the second adaptive weight vector at the moment; Get The sampling values ​​of the input signal at the moment and the previous multiple moments are used to construct the standard input signal vector; the standard input signal vector is spliced ​​with its conjugate vector to obtain the augmented input signal vector, and the obtained vector is constructed. The input matrix at time instant; based on The product of the input matrix at time t and the first adaptive weight vector and the second adaptive weight vector is calculated. The estimated signal at time t is used as a wide linear adaptive filter Output signal at the moment; calculate The difference between the expected signal and the estimated signal at the moment is obtained The estimated error signal at time t; based on The estimated error signal at the moment is updated using the minimum error modulus criterion to obtain Gain vector at time t; based on The gain vector at the moment and the estimated error signal are The first adaptive weight vector and the second adaptive weight vector at the moment are updated to obtain The first adaptive weight vector and the second adaptive weight vector at the moment reconstruct the wide linear adaptive filter The system weight vector at time t.

[0009] Preferably, the wide linear adaptive filter The system weight vector at time , is decomposed into tensors to obtain The first adaptive weight vector and the second adaptive weight vector at time t are expressed as: The first adaptive weight vector at time , expressed as: ; The second adaptive weight vector at time , expressed as: ; in, and Respectively The first adaptive weight vector at time and the second adaptive weight vector The sub-weight vectors, , represents the order of tensor decomposition, ; The length of each sub-weight vector in the first adaptive weight vector is , the length of each sub-weight vector in the second adaptive weight vector is , , is the length of the standard augmented complex adaptive filter weight vector; Represents a transpose operation.

[0010] Preferably, based on The estimated error signal at the moment is updated using the minimum error modulus criterion to obtain The gain vector at the moment includes: Based on the The length of each sub-weight vector in the adaptive weight vector is calculated The forgetting factor corresponding to the adaptive weight vector , expressed as: ; , represents the first adaptive weight vector, When represents the second adaptive weight vector; represents the preset hyperparameters, Indicates the The length of each sub-weight vector in the adaptive weight vector, represents the order of tensor decomposition; based on The input matrix and adaptive weight vector at the moment are used to construct the The adaptive weight vector corresponds to Regressor at time ,include: The first adaptive weight vector corresponds to The regressor at the moment is expressed as: ; Represents a transpose operation; The second adaptive weight vector corresponds to The regressor at the moment is expressed as: ; represents the conjugate transpose operation, express Input matrix at time The conjugate transpose of Based on the The adaptive weight vector corresponds to The product of the regression quantity at the moment and the inverse correlation matrix of the regression quantity is obtained The adaptive weight vector corresponds to Intermediate variables at the moment , expressed as: ; express Moment regressor The inverse correlation matrix of Based on the The adaptive weight vector corresponds to The intermediate variables, forgetting factors, estimated error signals and regression quantities at the moment are calculated. The adaptive weight vector corresponds to Gain vector at time , expressed as: ; express The estimated error signal at time t.

[0011] Preferably, based on The gain vector at the moment and the estimated error signal are The first adaptive weight vector and the second adaptive weight vector at the moment are updated to obtain The first adaptive weight vector and the second adaptive weight vector at the moment include: calculate The product of the gain vector at time t and the estimated error signal is The adaptive weight vectors at each moment are added to obtain The adaptive weight vector at the moment is expressed as: The first adaptive weight vector at time ; The second adaptive weight vector at time ; in, Represents the conjugate operation.

[0012] Preferably, the reconstruction wide linear adaptive filter The system weight vector at time t is expressed as: ; in, express The wide linear adaptive filter weight vector at time t; and Respectively The first adaptive weight vector at time and the second adaptive weight vector The sub-weight vectors; Represents the Kronecker product operation.

[0013] Preferably, based on The adaptive weight vector corresponds to Regressor at time The conjugate transpose of ,and The inverse correlation matrix , and The adaptive weight vector corresponds to The forgetting factor of the moment and gain vector , update to get the The adaptive weight vector corresponds to Regressor at time The inverse correlation matrix , expressed as: .

[0014] Preferably, The construction of the input matrix at each moment includes: Get Moment and its previous continuous The sampling value of the input signal at the moment , construct the standard input signal vector ; The standard input signal vector Its conjugate vector Splicing, get the length The augmented input signal vector ; Divide the augmented input signal vector into equal parts The length is The sub-input vector is obtained by equally dividing the augmented input signal vector, which is expressed as: ; Based on the equally divided augmented input signal vector, construct Input matrix at time ; in, represents the conjugate operation, Represents the conjugate transpose operation.

[0015] Preferably, based on the product of the input matrix and the first adaptive weight vector and the second adaptive weight vector, the calculation The estimated signal at time t is expressed as: ; in, express The estimated signal at time express The conjugate transpose vector of .

[0016] Preferably, calculate The difference between the expected signal and the estimated signal at the moment is obtained The estimated error signal at time t is expressed as: ; in, express The estimated error signal at time t, Indicates expected signal.

[0017] Preferably, when applied to a stereo video communication echo cancellation system, the method comprises: The input signal is a complex signal constructed by using two near-end speaker signals in a stereo video communication echo cancellation system as real and imaginary parts respectively; The expected signal is a complex signal constructed by using two near-end microphone signals in a stereo video communication echo cancellation system as the real part and the imaginary part respectively.

[0018] The above technical solution of the present invention has the following beneficial effects compared with the prior art:

[0019] The wide linear adaptive filter for pulse noise interference resistance described in the present invention utilizes a tensor integral decomposition to split the traditional high-dimensional system weight vector into two sets of low-dimensional adaptive weight vectors. This effectively reduces the computational effort required to update the inverse correlation matrix and improves the filter's tracking capability, thereby increasing the system's convergence speed and computational efficiency. Furthermore, the minimum error modulus criterion is introduced into the recursive adaptive algorithm, and a tensor integral decomposition method is employed to reduce the algorithm's latency and computational complexity, enabling robust adaptive filtering of the output signal of the system being estimated. Furthermore, the cost function selected by the present invention is based on the minimum error modulus at all times, resulting in faster convergence and more robust filtering performance when the desired signal is superimposed with pulse interference.

[0020] When applied to a stereo video communication echo cancellation system, the present invention utilizes a wide linear model to design an augmented complex adaptive filter, which can realize a dual-channel echo cancellation function with a single filter. The second-order non-circular characteristics of the complex signal can be fully utilized to improve the filtering capability, thereby achieving a better filtering effect and improving the communication quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to make the content 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, wherein: Figure 1 This is a working principle diagram of the wide linear adaptive filter for resisting impulse noise interference of the present invention; Figure 2 It is the system block diagram of the system identification model; Figure 3 It is a comparison chart of the normalized mean square deviation (NMSD) curve of the adaptive filtering system in the system identification scenario. DETAILED DESCRIPTION

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0023] Reference Figure 1 As shown in FIG. 1 , the working principle diagram of the wide linear adaptive filter for resisting impulse noise interference of the present invention includes: S101: Wide Linear Adaptive Filter The system weight vector at time , is decomposed into tensors to obtain The first adaptive weight vector and the second adaptive weight vector at the moment; S102: Acquisition The sampling values ​​of the input signal at the moment and the previous multiple moments are used to construct the standard input signal vector; the standard input signal vector is spliced ​​with its conjugate vector to obtain the augmented input signal vector, and the obtained vector is constructed. The input matrix at time instant; S103: Based on The product of the input matrix at time t and the first adaptive weight vector and the second adaptive weight vector is calculated. Estimated signal at time , as a wide linear adaptive filter The output signal at time is expressed as: ; in, express The conjugate transpose vector of ; S104: Calculation Expected signal at the moment The difference between the estimated signal and the The estimated error signal at time , expressed as: ; S105: Based on The estimated error signal at the moment is updated using the minimum error modulus criterion to obtain Gain vector at time t; S106: Based on The gain vector at the moment and the estimated error signal are The first adaptive weight vector and the second adaptive weight vector at the moment are updated to obtain The first adaptive weight vector and the second adaptive weight vector at the moment reconstruct the wide linear adaptive filter The system weight vector at time t.

[0024] In the embodiment of the present invention, the wide linear adaptive filter is divided into two parts, one of which is based on The input signal at the moment is filtered to obtain The estimated signal output at the moment; the other part is based on The estimated signal and the expected signal at the moment are used to calculate the estimated error signal. The system weight vector of the wide linear adaptive filter at the moment is updated to obtain A wide linear adaptive filter is used at all times to continuously filter and output the input signal.

[0025] Specifically, in step S101, the tensor integral decomposition wide linear adaptive filter The system weight vector at time , we get The first adaptive weight vector and the second adaptive weight vector at time t are expressed as: The first adaptive weight vector at time , expressed as: ; The second adaptive weight vector at time , expressed as: ; in, and Respectively The first adaptive weight vector at time and the second adaptive weight vector The sub-weight vectors, , represents the order of tensor decomposition, ; The length of each sub-weight vector in the first adaptive weight vector is , the length of each sub-weight vector in the second adaptive weight vector is , , is the length of the standard augmented complex adaptive filter weight vector; Represents a transpose operation.

[0026] The tensor integral decomposition method can improve the tracking speed of the adaptive filter for low-rank systems, reduce latency, and decompose the autocorrelation matrix in the adaptive algorithm into a smaller one, thereby saving computational effort. The low rank property here means that the optimal weights of the adaptive filter can be approximately arranged into a low-rank matrix. Therefore, the wide linear adaptive filter resistant to impulse noise interference of the present invention decomposes the long weight vector into a series of shorter weight vectors in the form of a sum of tensor products, which reduces the computational complexity of the adaptive filter and improves the tracking speed at an acceptable loss of accuracy.

[0027] Specifically, in step S102, the The input matrix at the moment includes: S102-1: Acquisition Moment and its previous continuous The sampling value of the input signal at the moment , construct the standard input signal vector ; S102-2: Input the standard signal vector Its conjugate vector Splicing, get the length The augmented input signal vector ;in, represents the conjugate operation, represents the conjugate transpose operation; S102-3: Divide the augmented input signal vector into equal parts The length is The sub-input vector is obtained by equally dividing the augmented input signal vector, which is expressed as: ; S102-4: Construct an input matrix based on the equally divided augmented input signal vector .

[0028] Specifically, in step S105, update the acquisition The gain vector at the moment includes: S105-1: Based on The length of each sub-weight vector in the adaptive weight vector is calculated The forgetting factor corresponding to the adaptive weight vector , expressed as: ; , represents the first adaptive weight vector, When represents the second adaptive weight vector; represents the preset hyperparameters, Indicates the The length of each sub-weight vector in the adaptive weight vector, represents the order of tensor decomposition; S105-2: Based on The input matrix and adaptive weight vector at the moment are used to construct the The adaptive weight vector corresponds to Regressor at time ,include: The first adaptive weight vector corresponds to The regressor at the moment is expressed as: ; Represents a transpose operation; The second adaptive weight vector corresponds to The regressor at the moment is expressed as: ; represents the conjugate transpose operation, express Input matrix at time The conjugate transpose of S105-3: Based on The adaptive weight vector corresponds to The product of the regression quantity at the moment and the inverse correlation matrix of the regression quantity is obtained The adaptive weight vector corresponds to Intermediate variables at the moment , expressed as: ; express Moment regressor The inverse correlation matrix of S105-4: Based on The adaptive weight vector corresponds to The intermediate variables, forgetting factors, estimated error signals and regression quantities at the moment are calculated. The adaptive weight vector corresponds to Gain vector at time , expressed as: ; express The estimated error signal at time t.

[0029] Among them, based on The adaptive weight vector corresponds to Regressor at time The conjugate transpose of ,and The inverse correlation matrix , and The adaptive weight vector corresponds to The forgetting factor of the moment and gain vector , update to get the The adaptive weight vector corresponds to Regressor at time The inverse correlation matrix , expressed as: .

[0030] Specifically, in step S106, calculate The product of the gain vector at time t and the estimated error signal is The adaptive weight vectors at each moment are added to obtain The adaptive weight vector at the moment is expressed as: The first adaptive weight vector at time ; The second adaptive weight vector at time .

[0031] Specifically, based on The adaptive weight vector at the moment reconstructs the wide linear adaptive filter The system weight vector at time t is expressed as: ; in, express The wide linear adaptive filter weight vector at time t; and Respectively The first adaptive weight vector at time and the second adaptive weight vector The sub-weight vectors; Represents the Kronecker product operation.

[0032] The wide linear adaptive filter for pulse noise interference resistance described in the present invention utilizes a tensor integral decomposition to split the traditional high-dimensional system weight vector into two sets of low-dimensional adaptive weight vectors. This effectively reduces the computational effort required to update the inverse correlation matrix and improves the filter's tracking capability, thereby increasing the system's convergence speed and computational efficiency. Furthermore, the minimum error modulus criterion is introduced into the recursive adaptive algorithm, and a tensor integral decomposition method is employed to reduce the algorithm's latency and computational complexity, enabling robust adaptive filtering of the output signal of the system being estimated. Furthermore, the cost function selected by the present invention is based on the minimum error modulus at all times, resulting in faster convergence and more robust filtering performance when the desired signal is superimposed with pulse interference.

[0033] Based on the above embodiment, in an embodiment of the present invention, when the wide linear adaptive filter for resisting impulse noise interference provided by the embodiment of the present invention is applied to an echo cancellation system for stereo video communication, the method includes: The input signal is a complex signal constructed by using two near-end speaker signals in a stereo video communication echo cancellation system as real and imaginary parts respectively; The expected signal is a complex signal constructed by using two near-end microphone signals in a stereo video communication echo cancellation system as the real part and the imaginary part respectively.

[0034] At this time, stereo video communication echo cancellation specifically includes: The two near-end speaker signals in the stereo video communication echo cancellation system are used as the real and imaginary parts respectively to construct the input signals of the wide linear adaptive filter in real time. The desired signal of the wide linear adaptive filter is constructed in real time using the two near-end microphone signals in the stereo video communication echo cancellation system as the real and imaginary parts respectively. The input signal is input into the wide linear adaptive filter as described above in real time to obtain Estimated signal at the moment; based on The estimated signal and expected signal at the moment are calculated The estimated error signal at time t is Pure signal output after echo elimination at all times; Using the minimum error modulus criterion, we can obtain Gain vector at time t; based on The gain vector at the moment and the estimated error signal are updated to obtain the wide linear adaptive filter System weight vector at time t; based on The system weight vector at the moment, as well as the input signal and expected signal input in real time, outputs a pure signal in real time, and updates the system weight vector of the wide linear adaptive filter at the next moment.

[0035] When applied to a stereo video communication echo cancellation system, the present invention utilizes a wide linear model to design an augmented complex adaptive filter, which can realize a dual-channel echo cancellation function with a single filter. The second-order non-circular characteristics of the complex signal can be fully utilized to improve the filtering capability, thereby achieving a better filtering effect and improving the communication quality.

[0036] Based on the above embodiment, in an embodiment of the present invention, using the wide linear adaptive filter with immunity to impulse noise interference provided by the present invention, in a scenario where the desired signal contains impulse noise, the steps of performing wide linear system identification specifically include: S201: Based on the system weight vector of the wide linear adaptive filter, two sets of adaptive weight vectors are constructed after tensor integral decomposition and ; Sub-weight vector and The lengths are and ,satisfy , is the length of the standard augmented complex adaptive filter weight vector; is the order of tensor decomposition, ; S202: By Moment and its previous continuous The sampled value of the complex input signal input to the system to be estimated , consisting of the standard input signal vector , concatenate the standard input signal vector and its conjugate vector to form , then the length is of Divide into The length is The sub-input vector of , and finally use each sub-input vector to form the input matrix ; superscript Represents the transpose operation; S203: By The two sets of adaptive weight vectors at the moment and the input matrix calculate the adaptive filter Estimated signal at time , expressed as: ; S204: Calculation The difference between the expected signal and the estimated signal at the moment is obtained The estimated error signal at time , expressed as: ; in, express The estimated error signal at time t, express The expected signal of the moment, express Estimated signal at the moment; S205: Based on the minimum error modulus criterion, update the adaptive weight vector and the inverse correlation matrix to obtain The tensor integral decomposition weight vector at time t includes: S205-1: Calculate the and exist Regressor at time 、 , and the forgetting factor 、 , expressed as: ; ; ; ; in, is a hyperparameter used to balance the convergence speed and steady-state error, , and They are and The length of the vector; Based on the minimum error modulus criterion, the present invention uses a recursive optimization method to derive the update formula of the tensor integral decomposition weight vector, which has a fast convergence speed and good robustness in environments containing impulse noise. S205-2: Calculate intermediate variables based on the inverse correlation matrix and regression 、 , expressed as: ; ; in, and They are After iterations and The inverse correlation matrix of S205-3: Calculation The gain vector at time t is expressed as: ; ; in, is the estimated error signal, and There are 2 forgetting factors, and They are used to update and The regressor of and are two intermediate variables; S205-4: Update The adaptive weight vector at the moment is expressed as: ; ; S205-5: Update The inverse correlation matrix at time is expressed as: ; ; During the iteration process, the initial value of the weight vector and Set as: , ,in, It means concatenating all the columns of a matrix into a column vector. Indicates size The semicolon means to concatenate the two matrices up and down. Indicates size The zero matrix of ; the initial value of the inverse correlation matrix 、 Set as 、 ; is a constant ,When the environmental noise is large, taking a smaller value is beneficial to the stability in the initial convergence stage.

[0037] S206: By Two sets of adaptive weight vectors of the time adaptive filter and , reconstruction The system weight vector at time , expressed as: ,in represents the Kronecker product operation; S207: By The input signal matrix at each moment is Time weight vector, calculate adaptive filter The estimated signal at time t is expressed as: .

[0038] The wide linear adaptive filter resistant to impulse noise interference of the present invention introduces the minimum error modulus criterion into the recursive adaptive algorithm and uses the tensor integral decomposition method to reduce the algorithm's delay and computational complexity, thereby achieving robust adaptive filtering of the output signal of the estimation system. The present invention utilizes a wide linear model to design an augmented complex adaptive filter, which can achieve dual-channel echo cancellation with a single filter. At the same time, the cost function selected by the present invention is based on the minimum error modulus at all times, which can obtain a filtering result with faster convergence speed and better robustness when the desired signal is superimposed with impulse interference. The tensor integral decomposition method utilized by the present invention decomposes long tap coefficients into the sum of tensor products of shorter weight vectors, which can effectively reduce the computational complexity of updating the inverse correlation matrix and improve the tracking capability of the filter.

[0039] To demonstrate the effectiveness of the present invention, this example uses computer experiments to verify the performance of the tensor integral decomposition-based wide linear recursive symbol (NKP-WL-RSA) adaptive filter provided by the present invention. The experiment estimates the unknown system in a system identification application scenario in an environment with impulse noise interference. The experimental results are compared with those of the wide linear recursive least squares (WL-RLS) adaptive filter, the tensor integral decomposition wide linear recursive least squares (NKP-WL-RLS) adaptive filter, and the wide linear recursive symbol (WL-RSA) adaptive filter.

[0040] Reference Figure 2 The figure shows the system block diagram of the system identification model. is Gaussian noise plus impulse noise; the system identification scene experiment uses normalized mean square deviation (NMSD) as the performance measure, that is, , in dB, where represents taking the logarithm, is the actual system weight.

[0041] In the experiment, the input signal is generated by passing a Gaussian white signal through a first-order autoregressive system with a pole of 0.9, and the noise signal is Contains a zero mean and a variance of Gaussian white noise and an impulse noise ,Right now Gaussian noise maintains a 30dB signal-to-noise ratio with the noise-free expected signal, while impulse noise maintains a -10dB signal-to-noise ratio with the noise-free expected signal. Produced by a Bernoulli Gaussian process, that is ,in is a Bernoulli process, and the probability of taking 0 is 0.9 and the probability of taking 1 is 0.1. is Gaussian white noise with zero mean.

[0042] Reference Figure 3 As shown in the figure, the normalized mean square deviation (NMSD) curve comparison of the adaptive filtering system in the system identification scenario; the parameters of each method are WL-RLS ( , ), NKP-WL-RLS( , , ), WL-RSA( , ), NKP-WL-RSA ( , , ).Depend on Figure 3 It can be seen that the NKP-WL-RSA adaptive filtering system of the embodiment of the present application has good anti-pulse convergence performance while maintaining low computational complexity, and can achieve faster convergence and tracking speed.

[0043] The wide linear adaptive filter for resisting impulse noise interference described in the present invention uses tensor integral decomposition to split the traditional high-dimensional system weight vector into two groups of low-dimensional adaptive weight vectors, which can effectively reduce the computational complexity of updating the inverse correlation matrix and improve the tracking ability of the filter, thereby improving the convergence speed and computational efficiency of the system; at the same time, the minimum error modulus criterion is introduced into the recursive adaptive algorithm, and the tensor integral decomposition method is used to reduce the algorithm's delay and reduce the computational complexity, thereby achieving robust adaptive filtering of the output signal of the estimated system. At the same time, the cost function selected by the present invention is based on the minimum error modulus at all times, which can obtain a filtering result with faster convergence speed and better robust performance when the desired signal is superimposed with impulse interference. When applied to a stereo video communication echo cancellation system, the present invention uses a wide linear model to design an augmented complex adaptive filter, which can realize the dual-channel echo cancellation function with a single filter, and can fully utilize the second-order non-circular characteristics of the complex signal to improve the filtering ability, thereby achieving better filtering effect and improving communication quality.

[0044] 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 take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] 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 block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0046] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0047] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device 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.

[0048] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A wide linear adaptive filter for resisting impulse noise interference, characterized in that: include: Wide Linear Adaptive Filter The system weight vector at time , is decomposed into tensors to obtain The first adaptive weight vector and the second adaptive weight vector at the moment; Get The sampling values ​​of the input signal at the moment and the previous multiple moments are used to construct the standard input signal vector; the standard input signal vector is spliced ​​with its conjugate vector to obtain the augmented input signal vector, and the obtained vector is constructed. The input matrix at time instant; based on The product of the input matrix at time t and the first adaptive weight vector and the second adaptive weight vector is calculated. The estimated signal at time t is used as a wide linear adaptive filter Output signal at the moment; calculate The difference between the expected signal and the estimated signal at the moment is obtained The estimated error signal at time t; based on The estimated error signal at the moment is updated using the minimum error modulus criterion to obtain Gain vector at time t; based on The gain vector at the moment and the estimated error signal are The first adaptive weight vector and the second adaptive weight vector at the moment are updated to obtain The first adaptive weight vector and the second adaptive weight vector at the moment reconstruct the wide linear adaptive filter The system weight vector at time t.

2. The wide linear adaptive filter for resisting impulse noise interference according to claim 1, characterized in that: Wide Linear Adaptive Filter The system weight vector at time , is decomposed into tensors to obtain The first adaptive weight vector and the second adaptive weight vector at time t are expressed as: The first adaptive weight vector at time , expressed as: ; The second adaptive weight vector at time , expressed as: ; in, and respectively The first adaptive weight vector representing the moment and the second adaptive weight vector The sub-weight vectors, , represents the order of tensor decomposition, ; The length of each sub-weight vector in the first adaptive weight vector is , the length of each sub-weight vector in the second adaptive weight vector is , , is the length of the standard augmented complex adaptive filter weight vector; Represents a transpose operation.

3. The wide linear adaptive filter for resisting impulse noise interference according to claim 2, characterized in that: based on The estimated error signal at the moment is updated using the minimum error modulus criterion to obtain The gain vector at the moment includes: Based on the The length of each sub-weight vector in the adaptive weight vector is calculated The forgetting factor corresponding to the adaptive weight vector , expressed as: ; , represents the first adaptive weight vector, When represents the second adaptive weight vector; represents the preset hyperparameters, Indicates the The length of each sub-weight vector in the adaptive weight vector, represents the order of tensor decomposition; based on The input matrix and adaptive weight vector at the moment are used to construct the The adaptive weight vector corresponds to Regressor at time ,include: The first adaptive weight vector corresponds to The regressor at the moment is expressed as: ; Represents a transpose operation; The second adaptive weight vector corresponds to The regressor at the moment is expressed as: ; represents the conjugate transpose operation, express Input matrix at time The conjugate transpose of Based on the The adaptive weight vector corresponds to The product of the regression quantity at the moment and the inverse correlation matrix of the regression quantity is obtained The adaptive weight vector corresponds to Intermediate variables at the moment , expressed as: ; express Moment regressor The inverse correlation matrix of Based on the The adaptive weight vector corresponds to The intermediate variables, forgetting factors, estimated error signals and regression quantities at the moment are calculated. The adaptive weight vector corresponds to Gain vector at time , expressed as: ; express The estimated error signal at time t.

4. The wide linear adaptive filter for resisting impulse noise interference according to claim 3, characterized in that: based on The gain vector at the moment and the estimated error signal are The first adaptive weight vector and the second adaptive weight vector at the moment are updated to obtain The first adaptive weight vector and the second adaptive weight vector at the moment include: calculate The product of the gain vector at time t and the estimated error signal is The adaptive weight vectors at each moment are added to obtain The adaptive weight vector at the moment is expressed as: The first adaptive weight vector at time ; The second adaptive weight vector at time ; in, Represents the conjugate operation.

5. The wide linear adaptive filter for resisting impulse noise interference according to claim 4, characterized in that: Reconstruction of wide linear adaptive filter The system weight vector at time t is expressed as: ； in, represents a wide linear adaptive filter System weight vector at time t; and Respectively The first adaptive weight vector at time and the second adaptive weight vector The sub-weight vectors; Represents the Kronecker product operation.

6. The wide linear adaptive filter for resisting impulse noise interference according to claim 3, characterized in that: Based on the The adaptive weight vector corresponds to Regressor at time The conjugate transpose of ,and The inverse correlation matrix , and The adaptive weight vector corresponds to The forgetting factor of the moment and gain vector , update to get the The adaptive weight vector corresponds to Regressor at time The inverse correlation matrix , expressed as: 。 7. The wide linear adaptive filter for resisting impulse noise interference according to claim 2, characterized in that: The construction of the input matrix at each moment includes: Get Moment and its previous continuous The sampling value of the input signal at the moment , construct the standard input signal vector ; The standard input signal vector Its conjugate vector Splicing, get the length The augmented input signal vector ; Divide the augmented input signal vector into equal parts The length is The sub-input vector is obtained by equally dividing the augmented input signal vector, which is expressed as: ; Based on the equally divided augmented input signal vector, construct Input matrix at time ; in, represents the conjugate operation, Represents the conjugate transpose operation.

8. The wide linear adaptive filter for resisting impulse noise interference according to claim 7, characterized in that: Based on the product of the input matrix and the first adaptive weight vector and the second adaptive weight vector, calculate The estimated signal at time t is expressed as: ； in, express The estimated signal at time express The conjugate transpose vector of .

9. The wide linear adaptive filter for resisting impulse noise interference according to claim 8, characterized in that: calculate The difference between the expected signal and the estimated signal at the moment is obtained The estimated error signal at time t is expressed as: ； in, express The estimated error signal at time t, express The expected signal of the moment.

10. The wide linear adaptive filter for resisting impulse noise interference according to claim 1, characterized in that: When used in stereo video communication echo cancellation systems, it includes: The input signal is a complex signal constructed by using two near-end speaker signals in a stereo video communication echo cancellation system as real and imaginary parts respectively; The expected signal is a complex signal constructed by using two near-end microphone signals in a stereo video communication echo cancellation system as the real part and the imaginary part respectively.

Images