A reverberation suppression method based on a time-frequency patch tensor model

Abstract

This application belongs to the field of underwater acoustic engineering technology. This application provides a reverberation suppression method based on a time-frequency patch tensor model. The embodiments of this disclosure utilize the low-rank sparsity characteristics of the echo signal in the time-frequency domain, modeling the reverberation suppression problem as an optimization problem of low-rank sparse decomposition. The time-frequency patch tensor is decomposed into a low-rank patch tensor and a sparse patch using an alternating direction multiplier method. The sparse patch tensor mainly includes the target echo component and a small portion of the reverberation component; the low-rank patch tensor mainly consists of the reverberation component. Since the low-rank patch tensor contains most of the reverberation component, the signal-to-mixing ratio of the obtained sparse patch tensor is significantly improved, thereby achieving reverberation suppression.

Description

Technical Field

[0001] This disclosure relates to the field of underwater acoustic engineering technology, and in particular to a reverberation suppression method based on a time-frequency patch tensor model. Background Technology

[0002] When using active sonar to detect underwater targets, reverberation interferes with the target echo signal. Reverberation is a random signal, generated by the random scattering and superposition of sound waves after encountering a large number of non-uniform scatterers underwater. Reverberation is one of the important factors affecting sonar detection performance, especially in shallow waters. How to effectively reduce the interference of reverberation on target detection has always been a research hotspot in the field of underwater acoustic signal processing.

[0003] Reverberation is strongly correlated with the transmitted signal and is colored and non-stationary, which limits the performance of conventional matched filters. In related technologies, a one-dimensional autoregressive pre-whitening method has been proposed. This method is a suboptimal signal detection method that assumes the local temporal stationarity of the reverberation background and uses the AR model coefficients of the current data segment to "whiten" the next data segment. However, due to the complexity of the underwater environment, the reverberation pre-whitening method may fail in environments with poor reverberation stationarity. Based on the power difference between reverberation and the signal, a reverberation suppression algorithm based on principal component inversion has also been proposed. This algorithm can effectively separate reverberation from the echo signal. Estimating the reverberation and signal power is crucial to the success of this method. However, the time-frequency characteristics of reverberation are related to the transmitted signal; in the time domain, reverberation overlaps with the target echo, while in the frequency domain, it is coherent with the target echo. It can be seen that removing reverberation solely from the time or frequency domain is difficult, necessitating the development of reverberation suppression methods that combine time and frequency domain information.

[0004] Therefore, it is necessary to improve one or more of the problems existing in the above-mentioned related technical solutions.

[0005] It should be noted that this section is intended to provide background or context for the technical solutions of this disclosure as set forth in the claims. The description herein does not constitute an admission that it is prior art simply because it is included in this section. Summary of the Invention

[0006] The purpose of this disclosure is to provide a reverberation suppression method based on a time-frequency patch tensor model, thereby overcoming at least to some extent one or more problems caused by the limitations and defects of related technologies.

[0007] According to embodiments of this disclosure, a reverberation suppression method based on a time-frequency patch tensor model is provided, the method comprising:

[0008] A short-time Fourier transform is performed on the echo data acquired in a single acquisition. The echo data after the short-time Fourier transform is then processed using modulo operation to obtain the time-frequency matrix and phase information.

[0009] Construct a time-frequency patch tensor based on the time-frequency matrix;

[0010] The time-frequency patch tensor is decomposed using the alternating direction multimultiplier method to obtain the sparse patch tensor;

[0011] The sparse patch tensor is reconstructed to obtain a sparse time-frequency matrix;

[0012] The phase information and the sparse time-frequency matrix are subjected to an inverse short-time Fourier transform to obtain the reverberation-suppressed time-domain data.

[0013] Further, the step of performing a short-time Fourier transform on the echo data acquired in a single acquisition, and processing the echo data after the short-time Fourier transform using modulo operations to obtain the time-frequency matrix and phase information includes:

[0014] For the echo signal of length L m Perform a short-time Fourier transform and obtain the time-frequency matrix using modulo operations. and the phase information ;in,

[0015] For the discretized echo signal m The expression for performing the short-time Fourier transform is:

[0016]

[0017] The window function uses the Hanning window, which is defined as follows:

[0018]

[0019] In the formula, w [ n [ ] represents the window function, with a length of M, n being the current window position, N being the number of points in the DFT, R being the number of overlap points, a being the frequency index, and b being the time index. = Let k be the time-frequency matrix, and k be the frequency point of the short-time Fourier transform. j The imaginary unit;

[0020] A represents the number of frequency-resolution elements, A = N / 2, and B represents the number of time-resolution elements. , This indicates rounding down to the nearest integer.

[0021] Furthermore, the step of constructing the time-frequency patch tensor based on the time-frequency matrix includes:

[0022] Use the sliding window from top left to bottom right A local patch matrix is ​​obtained by sliding with a preset sliding step size; where size is the dimension of the sliding window;

[0023] The patch matrices are stacked into the time-frequency patch tensor. .

[0024] Furthermore, the number of sliding windows for:

[0025]

[0026] In the formula, The sliding step size is in the vertical direction. This represents the sliding step size in the horizontal direction.

[0027] Further, the step of decomposing the time-frequency patch tensor using the alternating direction multimultiplier method to obtain the sparse patch tensor includes:

[0028] Let the low-rank patch tensor be The sparse patch tensor is Lagrange multipliers are The initial values ​​of the low-rank patch tensor, the sparse patch tensor, and the Lagrange multiplier are respectively set to... , and Let i = 1, 2, 3; let k be the iteration number, and set k = 0; the initial value of the penalty factor is... ;

[0029] Calculate the low-rank patch tensor for the k-th iteration; wherein, for the k-th iteration, the expression for calculating the low-rank patch tensor is:

[0030]

[0031] In the formula, This is the operation to restore the matrix expanded according to mode-i back to the original tensor. For the threshold shrinkage operator of matrix singular values, , For matrix SVD decomposition, ;

[0032] Based on the low-rank patch tensor of the k-th iteration, calculate the sparse patch tensor of the k-th iteration; wherein, the calculation expression of the sparse patch tensor of the k-th iteration is:

[0033]

[0034] Where λ is the weighting coefficient. For element-wise threshold shrinkage operator, , It is a symbolic function.

[0035] Based on the low-rank patch tensor and the sparse patch tensor of the k-th iteration, calculate the Lagrange multiplier of the k-th iteration; the expression for calculating the Lagrange multiplier in the k-th iteration is:

[0036]

[0037] Determine if the iteration stopping criterion is met:

[0038]

[0039]

[0040] in, It is a positive number;

[0041] If the iteration stopping criterion is not met, then update. ,renew The iteration is repeated; where, It is a constant;

[0042] If the stopping criterion described in the iteration is met, then output... , And stop the iteration process.

[0043] Further, the step of reconstructing the sparse patch tensor to obtain a sparse time-frequency matrix includes:

[0044] The value of each resolution cell is determined using a mean filtering function, the expression of which is:

[0045]

[0046] in, , A vector of values ​​corresponding to the p patch tensors of a single resolution unit;

[0047] The sparse time-frequency matrix is ​​obtained based on the value of each resolution unit. .

[0048] Further, the step of performing an inverse short-time Fourier transform on the phase information and the sparse time-frequency matrix to obtain the reverberation-suppressed time-domain data includes:

[0049] The phase information and the sparse time-frequency matrix are subjected to an inverse short-time Fourier transform, and the time-domain data is obtained using a synthesis window s. ;

[0050] The expression for the retrograde short-time Fourier transform is:

[0051]

[0052] The synthesis window s satisfies the energy complementarity constraint:

[0053] .

[0054] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects:

[0055] In the embodiments of this disclosure, the reverberation suppression method based on the time-frequency patch tensor model described above utilizes the low-rank sparsity characteristics of the echo signal in the time-frequency domain to model the reverberation suppression problem as an optimization problem of low-rank sparse decomposition. The time-frequency patch tensor is decomposed into a low-rank patch tensor and a sparse patch using the alternating direction multiplier method. The sparse patch tensor mainly includes the target echo component and a small portion of the reverberation component; the low-rank patch tensor mainly consists of the reverberation component. Since the low-rank patch tensor contains most of the reverberation component, the signal-to-mixing ratio of the resulting sparse patch tensor is significantly improved, thereby achieving reverberation suppression. Attached Figure Description

[0056] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0057] Figure 1 The diagram illustrates the steps of a reverberation suppression method based on a time-frequency patch tensor model in an exemplary embodiment of this disclosure.

[0058] Figure 2 A schematic diagram illustrating an experimental scenario in an exemplary embodiment of this disclosure is provided.

[0059] Figure 3 This illustrates beam domain echo data in an exemplary embodiment of this disclosure;

[0060] Figure 4 A detailed flowchart of the reverberation suppression method based on the time-frequency patch tensor model in an exemplary embodiment of this disclosure is shown.

[0061] Figure 5 This illustrates the time-frequency matrix obtained by short-time Fourier transform and modulus operation on beam domain echo data in an exemplary embodiment of this disclosure;

[0062] Figure 6 This illustrates the time-frequency matrix after reverberation suppression in an exemplary embodiment of this disclosure;

[0063] Figure 7 The image shows the signal after reverberation suppression using a reverberation suppression method based on a time-frequency patch tensor model, as illustrated in an exemplary embodiment of this disclosure.

[0064] Figure 8 In the exemplary embodiments shown in this disclosure Figure 3 The signal after matched filtering of beam domain echo data in the image;

[0065] Figure 9 In the exemplary embodiments shown in this disclosure Figure 5 The signal after reverberation suppression and then matched filtering is the signal in the process.

[0066] Figure 10 The diagram shows the receiving performance curves of the active sonar detection system corresponding to different reverberation suppression methods in the exemplary embodiments of this disclosure. Detailed Implementation

[0067] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0068] Furthermore, the accompanying drawings are merely illustrative diagrams of embodiments of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0069] This example implementation provides a reverberation suppression method based on a time-frequency patch tensor model. (See reference...) Figure 1 As shown, the reverberation suppression method based on the time-frequency patch tensor model may include steps S101 to S105.

[0070] Step S101: Perform a short-time Fourier transform on the echo data acquired in a single acquisition, and process the echo data after the short-time Fourier transform using modulo operation to obtain the time-frequency matrix and phase information;

[0071] Step S102: Construct a time-frequency patch tensor based on the time-frequency matrix;

[0072] Step S103: Decompose the time-frequency patch tensor using the alternating direction multimultiplier method to obtain the sparse patch tensor;

[0073] Step S104: Reconstruct the sparse patch tensor to obtain a sparse time-frequency matrix;

[0074] Step S105: Perform an inverse short-time Fourier transform on the phase information and the sparse time-frequency matrix to obtain the reverberation-suppressed time-domain data.

[0075] The reverberation suppression method based on the time-frequency patch tensor model described above utilizes the low-rank sparsity of the echo signal in the time-frequency domain to model the reverberation suppression problem as an optimization problem of low-rank sparse decomposition. The alternating direction multi-multiplier method is used to decompose the time-frequency patch tensor into a low-rank patch tensor and a sparse patch. The sparse patch tensor mainly includes the target echo component and a small portion of the reverberation component; the low-rank patch tensor mainly consists of the reverberation component. Since the low-rank patch tensor contains most of the reverberation component, the signal-to-mixing ratio of the resulting sparse patch tensor is significantly improved, thus achieving reverberation suppression.

[0076] Below, we will refer to Figures 1 to 10 The reverberation suppression method based on the time-frequency patch tensor model described in this example embodiment will be explained in more detail.

[0077] In step S101, a short-time Fourier transform (STFT) is performed on the echo data m acquired in a single acquisition, and then a time-frequency matrix is ​​constructed using modulo operations. And save phase information ;

[0078] For the discretized echo signal m, its time spectrum is:

[0079] (1)

[0080] In the formula, w is the window function with length M, N is the number of points in the DFT, R is the number of overlapping points, a is the frequency index, and b is the time index. In the method proposed in this application, the Hanning window is used, and its definition is as follows:

[0081] (2)

[0082] For an echo signal m of length L, perform STFT and obtain the time-frequency matrix using modulo operation. and phase Where A is the number of frequency-resolution units, A = N / 2; and B is the number of time-resolution units. , This indicates rounding down to the nearest integer.

[0083] In step S102, the time-frequency matrix is ​​used. Constructing the time-frequency patch tensor From the time-frequency matrix Constructing the time-frequency patch tensor The specific steps are as follows:

[0084] (1) Using a sliding window This yields a series of local patch matrices, where size is the dimension of the sliding window. From the time-frequency matrix Starting from the top left corner with sliding steps Slide horizontally until it can no longer move horizontally; then the window returns to the far left and moves in increments. Move vertically downwards. Then, repeat this process until the entire time-frequency matrix has been traversed.

[0085] (2) These patches are stacked into a three-dimensional cube, which this application refers to as the patch tensor. ;

[0086] in The number of sliding windows.

[0087] (3)

[0088] In this way, the time-frequency matrix can be... Convert to time-frequency patch tensor .

[0089] In step S103, the time-frequency patch tensor is processed using the alternating direction multimultiplier method. Decomposed into low-rank patch tensors and sparse patch tensors The specific decomposition process is as follows:

[0090] (1) Variable initialization;

[0091] Let tensor For low-rank patch tensors, tensors For sparse patch tensors, tensors It is a Lagrange multiplier; tensor tensor as well as The initial values ​​are respectively , and Let i = 1, 2, 3; let k be the iteration number, and set k = 0; the initial value of the penalty factor is... .

[0092] (2) Calculate the low-rank patch tensor of the k-th iteration; the expression for calculating the low-rank patch tensor of the k-th iteration is:

[0093] (4)

[0094] in, For the threshold shrinkage operator of matrix singular values, ,in For matrix SVD decomposition, . This is the operation of restoring the matrix expanded according to mode-i back to the original tensor.

[0095] (3) Calculate the sparse patch tensor for the k-th iteration; the expression for calculating the sparse patch tensor for the k-th iteration is:

[0096] (5)

[0097] Where λ is the weighting coefficient, which controls the sparsity of the target patch tensor; This is an element-wise threshold shrinkage operator. For a given shrinkage threshold... , , It is a symbolic function.

[0098] (4) Calculate the Lagrange multipliers for the k-th iteration; the expression for calculating the Lagrange multipliers for the k-th iteration is:

[0099] (6)

[0100] (5) Determine if the condition for stopping iteration has been reached; determine if the algorithm meets the iteration stopping criterion:

[0101] and (7)

[0102] in It is a small positive number, depending on the precision required by the algorithm.

[0103] If the iteration stopping criterion is not met, then update. ,renew

[0104] (8)

[0105] Then repeat the process (2)-(4) above until the algorithm iteration stopping criterion is met.

[0106] If the iteration stopping criterion is met, then output... , And stop the iteration process.

[0107] In step S104, from the sparse patch tensor Sparse time-frequency matrix reconstructed from From sparse patch tensors Sparse time-frequency matrix reconstructed from The specific steps are as follows:

[0108] Determining the time-frequency matrix using the mean filter function The value of each resolution unit, the time-frequency matrix The value of each resolving unit corresponds to the value from the sparse patch tensor. Several values ​​are given. Therefore, the mean filter function is used to determine the value of the resolving unit. :

[0109] (9)

[0110] in , It is a vector containing the values ​​of the p patch tensors corresponding to a single resolution unit. The time-frequency matrix is ​​determined using the mean filtering function. Each resolution unit value yields a sparse time-frequency matrix. .

[0111] In step S105, phase information is used and sparse time-frequency matrix After performing inverse STFT, the reverberation-suppressed time-domain data is obtained. This is the result after reverberation suppression.

[0112] The inversion of STFT is performed by the synthesis operator. To achieve this, the signal is obtained using a synthesis window s. :

[0113] (10)

[0114] The synthesis window s satisfies the energy complementarity constraint:

[0115] (11)

[0116] In one specific embodiment, the present invention will be further described below with reference to the accompanying drawings and embodiments. The present invention includes, but is not limited to, the following embodiments.

[0117] 1. Experimental setup.

[0118] Experimental scenarios, such as Figure 2 As shown, the active sonar was fixed at a depth of 40 m, and the target was a small sphere. The active detection system transmitted signals at 1 s intervals, using LFM signals with a normalized frequency range of 0.56–0.65 Hz. The receiver array was a uniform linear array of 64 elements with a spacing of 3 cm between elements, a sampling rate of 100 kHz, and an effective horizontal field of view of 180 degrees. The experiment was conducted in a lake environment, where the main interference was reverberation.

[0119] 2. Experimental data.

[0120] The active detection system repeatedly transmits linear frequency modulated signals, and the echo data acquired by the linear array is processed by beamforming to obtain beam domain data, such as... Figure 3 As shown.

[0121] like Figure 4 As shown, the specific implementation process of the reverberation suppression method based on the time-frequency patch tensor model proposed in this invention is as follows:

[0122] A short-time Fourier transform (STFT) is performed on the beam domain echo data m acquired in a single acquisition, and then a time-frequency matrix is ​​constructed using modulo operations. And save phase information ;

[0123] For the discretized echo signal m, its time spectrum is:

[0124]

[0125] Where w is the Hanning window with a length of 64, an overlap of 32, a DFT points of 128, a is the frequency index, and b is the time index.

[0126] For an echo signal m of length 10000, perform STFT and obtain the time-frequency matrix using modulo operation. ,like Figure 5 As shown.

[0127] Using time-frequency matrix Constructing the time-frequency patch tensor From the time-frequency matrix Constructing the time-frequency patch tensor The specific steps are as follows:

[0128] (1) Use a sliding window from the top left to the bottom right With horizontal sliding step size Slide until it can no longer move horizontally; then the window returns to the far left and moves in increments. Move vertically downwards. Then repeat this process until the entire time-frequency matrix has been traversed. A series of local patch matrices are obtained by sliding with a certain step size.

[0129] (2) These patches are stacked into a three-dimensional cube, which this application refers to as the patch tensor. ;

[0130] In this way, the time-frequency matrix can be... Convert to time-frequency patch tensor .

[0131] Using the alternating direction multimultiplier method to transform the time-frequency patch tensor Decomposed into low-rank patch tensors and sparse patch tensors The specific decomposition process is as follows:

[0132] (1) Variable initialization;

[0133] Let tensor For low-rank patch tensors, tensors For sparse patch tensors, tensors It is a Lagrange multiplier; tensor tensor as well as The initial values ​​are respectively , and Let i = 1, 2, 3; let k be the iteration number, and set k = 0; the initial value of the penalty factor is... .

[0134] (2) Calculate the low-rank patch tensor of the k-th iteration; the expression for calculating the low-rank patch tensor of the k-th iteration is:

[0135]

[0136] in For the threshold shrinkage operator of matrix singular values, ,in For matrix SVD decomposition, . This is the operation of restoring the matrix expanded according to mode-i back to the original tensor.

[0137] (3) Calculate the sparse patch tensor for the k-th iteration; the expression for calculating the sparse patch tensor for the k-th iteration is:

[0138]

[0139] Where λ is the weighting coefficient, taken as... ; This is an element-wise threshold shrinkage operator. For a given shrinkage threshold... , , It is a symbolic function.

[0140] (4) Calculate the Lagrange multipliers for the k-th iteration; the expression for calculating the Lagrange multipliers for the k-th iteration is:

[0141]

[0142] (5) Determine if the condition for stopping iteration has been reached; determine if the algorithm meets the iteration stopping criterion:

[0143] and

[0144] If the iteration stopping criterion is not met, then update. ,renew

[0145]

[0146] Then repeat the process (2)-(4) above until the algorithm iteration stopping criterion is met.

[0147] If the iteration stopping criterion is met, then output... , And stop the iteration process.

[0148] From sparse patch tensor Sparse time-frequency matrix reconstructed from From sparse patch tensors Sparse time-frequency matrix reconstructed from The steps are as follows:

[0149] Determining the time-frequency matrix using the mean filter function The value of each resolution unit, the time-frequency matrix The value of each resolving unit corresponds to the value from the sparse patch tensor. Several values ​​are given. Therefore, the mean filter function is used to determine the value of the resolving unit. :

[0150]

[0151] in , It is a vector containing the values ​​of the p patch tensors corresponding to a single resolution unit. The time-frequency matrix is ​​determined using the mean filtering function. Each resolution unit value yields the time-frequency matrix. ,like Figure 6 As shown.

[0152] Using phase information and sparse time-frequency matrix After performing inverse STFT, the reverberation-suppressed time-domain data is obtained. This is the result after reverberation suppression.

[0153] The final result after reverberation suppression is as follows: Figure 7 As shown. To compare the reverberation suppression effect, a matched filter was chosen as the detector. Figure 3 The beam domain data shown is the result of matched filtering. Figure 8 As shown, Figure 7 The result of the reverberation-suppressed signal after matched filtering is shown below. Figure 9 As shown. Comparison Figure 8 and Figure 9 It can be seen that the amplitude of reverberation interference can be significantly suppressed after processing by the method proposed in this invention.

[0154] To verify the innovativeness of the method proposed in this invention, simulation data was used to compare it with reverberation pre-whitening and principal component inversion methods. The results are as follows: Figure 10 As shown, under different signal-to-mixing ratios, the method proposed in this invention achieves better reverberation suppression than the reverberation pre-whitening method and the principal component inversion method, thus improving the detection performance of active sonar.

[0155] The reverberation suppression method based on the time-frequency patch tensor model described above utilizes the low-rank sparsity of the echo signal in the time-frequency domain to model the reverberation suppression problem as an optimization problem of low-rank sparse decomposition. The alternating direction multi-multiplier method is used to decompose the time-frequency patch tensor into a low-rank patch tensor and a sparse patch. The sparse patch tensor mainly includes the target echo component and a small portion of the reverberation component; the low-rank patch tensor mainly consists of the reverberation component. Since the low-rank patch tensor contains most of the reverberation component, the signal-to-mixing ratio of the resulting sparse patch tensor is significantly improved, thus achieving reverberation suppression.

[0156] It should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc., in the above description indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this disclosure and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this disclosure.

[0157] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0158] In the embodiments of this disclosure, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure according to the specific circumstances.

[0159] In embodiments of this disclosure, unless otherwise expressly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0160] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0161] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A reverberation suppression method based on a time-frequency patch tensor model, characterized in that, The method includes: A short-time Fourier transform is performed on the echo data acquired in a single acquisition. The echo data after the short-time Fourier transform is then processed using modulo operation to obtain the time-frequency matrix and phase information. Construct a time-frequency patch tensor based on the time-frequency matrix; The time-frequency patch tensor is decomposed using the alternating direction multimultiplier method to obtain the sparse patch tensor; The sparse patch tensor is reconstructed to obtain a sparse time-frequency matrix; The phase information and the sparse time-frequency matrix are subjected to an inverse short-time Fourier transform to obtain the reverberation-suppressed time-domain data.

2. The reverberation suppression method based on the time-frequency patch tensor model according to claim 1, characterized in that, The steps of performing a short-time Fourier transform on the echo data acquired in a single acquisition, and processing the transformed echo data using modulo operations to obtain the time-frequency matrix and phase information, include: For echo signals of length L m Perform a short-time Fourier transform and obtain the time-frequency matrix using modulo operations. and the phase information ;in, For the discretized echo signal m The expression for performing the short-time Fourier transform is: The window function uses the Hanning window, which is defined as follows: In the formula, w Let be the window function, with length M, n be the current window position, N be the number of points in the DFT, R be the number of overlap points, a be the frequency index, and b be the time index. = It is a time-frequency matrix. j The imaginary unit; A represents the number of frequency-resolution elements, A = N / 2, and B represents the number of time-resolution elements. , This indicates rounding down to the nearest integer.

3. The reverberation suppression method based on the time-frequency patch tensor model according to claim 2, characterized in that, The step of constructing a time-frequency patch tensor based on the time-frequency matrix includes: Use the sliding window from top left to bottom right A local patch matrix is ​​obtained by sliding with a preset sliding step size; where size is the dimension of the sliding window; The patch matrices are stacked into the time-frequency patch tensor. .

4. The reverberation suppression method based on the time-frequency patch tensor model according to claim 3, characterized in that, The number of sliding windows for: In the formula, The sliding step size is in the vertical direction. This represents the sliding step size in the horizontal direction.

5. The reverberation suppression method based on the time-frequency patch tensor model according to claim 4, characterized in that, The step of decomposing the time-frequency patch tensor using the alternating direction multimultiplier method to obtain the sparse patch tensor includes: Let the low-rank patch tensor be The sparse patch tensor is Lagrange multipliers are The initial values ​​of the low-rank patch tensor, the sparse patch tensor, and the Lagrange multiplier are respectively set to... , and Let i = 1, 2, 3; let k be the iteration number, and set k = 0; the initial value of the penalty factor is... , For time-frequency patch tensor Vectorized standard deviation; Calculate the low-rank patch tensor for the k-th iteration; wherein, for the k-th iteration, the expression for calculating the low-rank patch tensor is: In the formula, This is the operation to restore the matrix expanded according to mode-i back to the original tensor. For the threshold shrinkage operator of matrix singular values, , For matrix SVD decomposition, It is a matrix containing A vector of singular values, for After threshold shrinkage operator The result after processing As a penalty factor, Let Lagrange multipliers be the multipliers in the k-th iteration. Let be the sparse patch tensor of the k-th iteration; Based on the low-rank patch tensor of the k-th iteration, calculate the sparse patch tensor of the k-th iteration; wherein, the calculation expression of the sparse patch tensor of the k-th iteration is: Where λ is the weighting coefficient. For element-wise threshold shrinkage operator, , It is a symbolic function; Based on the low-rank patch tensor and the sparse patch tensor of the k-th iteration, calculate the Lagrange multiplier of the k-th iteration; the expression for calculating the Lagrange multiplier in the k-th iteration is: Determine if the iteration stopping criterion is met: in, It is a positive number; If the iteration stopping criterion is not met, then update. ,renew The iteration is repeated; where, It is a constant; If the stopping criterion described in the iteration is met, then output... , And stop the iteration process.

6. The reverberation suppression method based on the time-frequency patch tensor model according to claim 5, characterized in that, The step of reconstructing the sparse patch tensor to obtain a sparse time-frequency matrix includes: The value of each resolution cell is determined using a mean filtering function, the expression of which is: in, , A vector of values ​​corresponding to the p patch tensors of a single resolution unit; The sparse time-frequency matrix is ​​obtained based on the value of each resolution unit. .

7. The reverberation suppression method based on the time-frequency patch tensor model according to claim 6, characterized in that, The step of performing an inverse short-time Fourier transform on the phase information and the sparse time-frequency matrix to obtain the reverberation-suppressed time-domain data includes: The phase information and the sparse time-frequency matrix are subjected to an inverse short-time Fourier transform, and the time-domain data is obtained using a synthesis window s. ; The expression for the retrograde short-time Fourier transform is: The synthesis window s satisfies the energy complementarity constraint: in, It is a time-frequency matrix. Let be the value of the window function s at n-bR. For window functions The value at n+bR Let be the value of the window function s at n+bR.

Images