A signal blind extraction method, device, equipment and medium without error accumulation
Through the gradual progressive blind extraction method, error accumulation is avoided, signal extraction accuracy is improved, error propagation problem in sequential blind signal extraction is solved, and high-precision source signal extraction is achieved.
Patent Information
- Application Number
- CN202411100748.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-12
AI Technical Summary
There is an error accumulation problem in the sequential blind signal extraction algorithm, which leads to a gradual decrease in estimation accuracy.
The blind extraction vector is obtained by a stepwise progressive method, the first blind extraction vector is estimated by the covariance matrix of the observed signal vector, and the blind extraction cost function is constructed based on the blind extraction criterion, and the optimal solution is searched using the gradient descent optimization algorithm, and the optimal solution is searched. It is projected to a subspace orthogonal to the vector of the previous step to avoid error propagation until it is an iterative search to obtain all blindly extracted vectors.
It effectively eliminates the accumulation of errors during the blind extraction process, improves the accuracy of source signal extraction, and maintains the flexibility of the sequential blind extraction algorithm.
Smart Images

Figure CN118643319B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of signal processing technology, and in particular to a method, device, equipment and medium for blind signal extraction without error accumulation. Background Art
[0002] Blind Signal Extraction (BSE) is a special type of blind separation method, which is widely used in electronic information, biomedicine, radar / sonar signal processing, wireless communication, fault diagnosis and other technical fields. Blind signal extraction extracts source signals one by one in a sequential manner, rather than recovering all source signals at once. Compared with simultaneous blind separation (Simultaneous Blind Separation), it has two advantages: First, the blind extraction method has better flexibility, because it can only extract the source signal of interest according to the actual application situation, so it is more economical; especially when the number of source signals is large and the source signals to be extracted only account for a small part of them, this advantage will be more prominent. Second, the blind extraction algorithm is usually simpler in form and has lower computational complexity.
[0003] However, the sequential blind extraction algorithm relies on the deflation procedure, and the traditional deflation procedure suffers from the problem of error accumulation. This is because the estimation error generated in the previous estimation step during the deflation process will be propagated to the subsequent estimation step, affecting the accuracy of the estimation result in the subsequent estimation process, thereby causing the estimation accuracy to gradually decrease. Summary of the invention
[0004] Based on this, it is necessary to provide a signal blind extraction method, device, equipment and medium without error accumulation, which can eliminate the error accumulation in the compression process of signal blind extraction and improve the accuracy of source signal extraction, in order to solve the above technical problems.
[0005] A signal blind extraction method without error accumulation, the method comprising:
[0006] Get the observed signal vector; where the observed signal vector contains the M Source signal;
[0007] Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix;
[0008] According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector;
[0009] The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is reconstructed based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching;
[0010] The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors;
[0011] All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0012] In one embodiment, obtaining an observation signal vector includes:
[0013] Get The observed signal vector at time t is expressed as ,in, No. Observed signals, superscript represents transpose, is the number of observed signals in the observed signal vector, and the observed signal vector contains Source signals to be extracted .
[0014] In one embodiment, estimating a first blind extraction vector according to a covariance matrix of an observed signal vector, and using the first blind extraction vector as a first row vector in a blind extraction matrix, comprises:
[0015] When the calculation time delay is 0, The observed signal vector at time The covariance matrix of ;
[0016] According to the covariance matrix The matrix consisting of the left singular vectors of , further calculations yield the matrix and matrix ;in, Indicates the time delay is hour, The covariance matrix of Indicates the time delay is hour, The covariance matrix of represents the number of covariance matrices;
[0017] Among them, the matrix and matrix There are several generalized eigenvalues, and the generalized eigenvector corresponding to the largest generalized eigenvalue is used as the first blind extraction vector , and the first blind extraction vector As a blind extraction matrix The first row vector in .
[0018] In one embodiment, a blind extraction cost function of the observed signal vector is constructed according to the blind extraction vector and based on the blind extraction criterion, which is expressed as: ;in, is the blind extraction vector, the superscript represents transpose, is the matrix and matrix Construct the intermediate matrix obtained, and The row vector of , Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors; .
[0019] In one embodiment, the blind extraction criterion is expressed as: for the blind extraction vector , and parallel.
[0020] In one embodiment, a gradient descent optimization algorithm is used to search for an optimal solution of a blind extraction cost function, including:
[0021] Using gradient descent optimization algorithm to search blind extraction cost function The optimal solution ; Among them, the gradient descent optimization algorithm is expressed as ;in, Represents the blind extraction vector The Frobenius norm of , is the learning rate, Blind extraction vector The iterative update rule is: Represents the blind extraction vector It plays a normalizing role.
[0022] In one embodiment, performing optimal solution projection according to an exact solution subspace search method includes:
[0023] For the previously acquired Blind extraction vector Transpose to obtain the transposed blind extraction vector , and through Perform singular value decomposition, corresponding to the singular value 0 Left singular vectors The space formed is the same as the previous one Blind extraction vector orthogonal subspace; where is the number of source signals to be extracted hidden in the observed signal vector;
[0024] According to the exact solution subspace search method, the blind extraction cost function The optimal solution Project to the previous Blind extraction vector In the orthogonal subspace, the optimal solution of the blind extraction cost function is not equal to the estimated prior Any one of the blind extraction vectors;
[0025] judge Whether the iterative process converges, if not, reconstruct the blind extraction cost function and solve the optimal solution; if converged, according to the subspace formed by all optimal solutions of the blind extraction cost function and does not contain the previous Blind extraction vector Subspace of The intersection of Blind extraction vector .
[0026] A signal blind extraction device without error accumulation, the device comprising:
[0027] The data acquisition module is used to obtain the observation signal vector; wherein the observation signal vector contains the data to be extracted. M Source signal;
[0028] A first blind extraction vector estimation module is used to estimate a first blind extraction vector according to a covariance matrix of an observed signal vector, and use the first blind extraction vector as a first row vector in a blind extraction matrix;
[0029] The blind extraction vector iterative search module is used to construct a blind extraction cost function of the observed signal vector based on the first blind extraction vector and the blind extraction criterion, use the gradient descent optimization algorithm to search for the optimal solution of the blind extraction cost function, and project the optimal solution to a subspace orthogonal to the first blind extraction vector according to the exact solution subspace search method, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and search to obtain the second blind extraction vector; the second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is constructed again based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected again to a subspace orthogonal to both the first blind extraction vector and the second blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and search to obtain the third blind extraction vector; the third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is constructed repeatedly, the optimal solution is obtained by optimization, and the optimal solution is projected according to the exact solution subspace search method, until the iterative search obtains the third blind extraction vector. M Blind extraction vectors;
[0030] The source signal extraction module is used to use all the blind extraction vectors obtained above as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0031] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0032] Get the observed signal vector; where the observed signal vector contains the M Source signal;
[0033] Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix;
[0034] According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector;
[0035] The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is reconstructed based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching;
[0036] The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors;
[0037] All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0038] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:
[0039] Get the observed signal vector; where the observed signal vector contains the M Source signal;
[0040] Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix;
[0041] According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector;
[0042] The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is reconstructed based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching;
[0043] The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors;
[0044] All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0045] The above-mentioned method, device, equipment and medium for blind signal extraction without error accumulation adopts a step-by-step approach to search and obtain blind extraction vectors. For the first blind extraction vector, it is estimated based on the covariance matrix of the observed signal vector; for other blind extraction vectors, a blind extraction cost function is constructed based on the blind extraction vectors that have been searched and obtained, the optimal solution is obtained by searching and optimizing, and the optimal solution is projected based on the exact solution subspace search method. By projecting the optimal solution into a subspace that is orthogonal to the blind extraction vectors that have been searched and obtained, it is restricted that the optimal solution is not equal to the blind extraction vector that has been searched and obtained, until the iterative search obtains the first blind extraction vector. M The blind extraction vector is obtained, so that the accuracy of the next blind extraction vector is not affected by the error of the blind extraction vector that has been searched and obtained, avoiding the accumulation of errors in the blind extraction process. Finally, all the blind extraction vectors obtained are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The method can effectively improve the accuracy of source signal extraction by eliminating the error accumulation in the blind extraction process while maintaining the flexibility of the sequential blind extraction algorithm. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 It is a flowchart of a signal blind extraction method without error accumulation in one embodiment;
[0047] Figure 2 A schematic diagram of a process of iteratively searching for blind extraction vectors in one embodiment;
[0048] Figure 3 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0050] In one embodiment, Figure 1 As shown, a signal blind extraction method without error accumulation is provided, comprising the following steps:
[0051] Step S1, obtaining an observation signal vector; wherein the observation signal vector contains the M source signal.
[0052] It can be understood that the observation signals include communication signals, radar echo signals, optical signals, and electrical signals, etc.
[0053] Step S2, estimating a first blind extraction vector according to the covariance matrix of the observed signal vector, and using the first blind extraction vector as the first row vector in the blind extraction matrix.
[0054] Step S3, constructing a blind extraction cost function of the observed signal vector according to the first blind extraction vector and based on the blind extraction criterion, using the gradient descent optimization algorithm to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, projecting the optimal solution into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain the second blind extraction vector.
[0055] Step S4, taking the second blind extraction vector as the second row vector in the blind extraction matrix, and re-constructing the blind extraction cost function of the observed signal vector based on the first blind extraction vector and the second blind extraction vector and based on the blind extraction criterion, and again using the gradient descent optimization algorithm to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, again projecting the optimal solution into a subspace that is orthogonal to both the first blind extraction vector and the second blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and searching to obtain the third blind extraction vector.
[0056] Step S5, taking the third blind extraction vector as the third row vector in the blind extraction matrix, and repeatedly constructing the blind extraction cost function, optimizing to obtain the optimal solution, and performing optimal solution projection according to the exact solution subspace search method, until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vector.
[0057] Step S6: All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0058] In one embodiment, obtaining an observation signal vector includes:
[0059] Get The observed signal vector at time t is expressed as ,in, No. Observed signals, superscript represents transpose, is the number of observed signals in the observed signal vector, and the observed signal vector contains Source signals to be extracted .
[0060] In one embodiment, estimating a first blind extraction vector according to a covariance matrix of an observed signal vector, and using the first blind extraction vector as a first row vector in a blind extraction matrix, comprises:
[0061] When the calculation time delay is 0, The observed signal vector at time The covariance matrix of ;
[0062] By changing the covariance matrix Perform singular value decomposition according to the covariance matrix The matrix consisting of the left singular vectors of , further calculations yield the matrix and matrix ;in, Indicates the time delay is hour, The covariance matrix of Indicates the time delay is hour, The covariance matrix of represents the number of covariance matrices;
[0063] Among them, the matrix and matrix There are several generalized eigenvalues, and the generalized eigenvector corresponding to the largest generalized eigenvalue is used as the first blind extraction vector , and the first blind extraction vector As a blind extraction matrix The first row vector in .
[0064] In one embodiment, a blind extraction cost function of the observed signal vector is constructed according to the blind extraction vector and based on the blind extraction criterion, which is expressed as: ;in, is the blind extraction vector, the superscript T represents transpose, is the matrix and matrix Construct the intermediate matrix obtained, and The row vector of , Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors; .
[0065] In one embodiment, the blind extraction criterion is expressed as: for the blind extraction vector , and parallel.
[0066] In one embodiment, a gradient descent optimization algorithm is used to search for an optimal solution of a blind extraction cost function, including:
[0067] Using gradient descent optimization algorithm to search blind extraction cost function The optimal solution ; Among them, the gradient descent optimization algorithm is expressed as ;in, Represents the blind extraction vector The Frobenius norm of , is the learning rate, Blind extraction vector The iterative update rule is: Represents the blind extraction vector It plays a role of normalization. Among them, the Frobenius norm, referred to as F norm, is a matrix norm. For a matrix A, its Frobenius norm is defined as the square root of the sum of the squares of all elements of the matrix A.
[0068] In one embodiment, the process of iteratively searching for blind extraction vectors is as follows: Figure 2 As shown, in the process of iteratively searching for blind extraction vectors, optimal solution projection is performed according to the exact solution subspace search method, including:
[0069] For the previously acquired Blind extraction vector Transpose to obtain the transposed blind extraction vector , and through Perform singular value decomposition, corresponding to the singular value 0 Left singular vectors The space formed is the same as the previous one Blind extraction vector orthogonal subspace; where is the number of source signals to be extracted hidden in the observed signal vector;
[0070] According to the exact solution subspace search method, the blind extraction cost function The optimal solution Project to the previous Blind extraction vector In the orthogonal subspace, the optimal solution of the blind extraction cost function is not equal to the estimated prior Any one of the blind extraction vectors;
[0071] judge Whether the iterative process converges, if not, reconstruct the blind extraction cost function and solve the optimal solution; if converged, according to the subspace formed by all optimal solutions of the blind extraction cost function and does not contain the previous Blind extraction vector Subspace of The intersection of Blind extraction vector .
[0072] It can be understood that the principle of exact solution subspace search is: express The subspace composed of all optimal solutions is represents another subspace (which does not contain the previous Blind extraction vector ),So belong and The intersection of . The purpose is to prevent the iterative search process from converging to the previous Blind extraction vector . is called an “exact solution subspace search” because the space contains At the same time By excluding it, the accuracy of the blind extraction vector currently obtained by search is not affected by the error of the blind extraction vector that has been obtained, thereby eliminating the error accumulation in the blind extraction process.
[0073] Furthermore, all the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0074] Specifically, based on The source signal extraction result obtained by the row vector is expressed as , based on the first to The source signal extraction result obtained by the row vector is expressed as: .
[0075] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.
[0076] In one embodiment, a signal blind extraction device without error accumulation is provided, comprising:
[0077] The data acquisition module is used to obtain the observation signal vector; wherein the observation signal vector contains the data to be extracted. M Source signal;
[0078] A first blind extraction vector estimation module is used to estimate a first blind extraction vector according to a covariance matrix of an observed signal vector, and use the first blind extraction vector as a first row vector in a blind extraction matrix;
[0079] The blind extraction vector iterative search module is used to construct a blind extraction cost function of the observed signal vector based on the first blind extraction vector and the blind extraction criterion, use the gradient descent optimization algorithm to search for the optimal solution of the blind extraction cost function, and project the optimal solution to a subspace orthogonal to the first blind extraction vector according to the exact solution subspace search method, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and search to obtain the second blind extraction vector; the second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is constructed again based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected again to a subspace orthogonal to both the first blind extraction vector and the second blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and search to obtain the third blind extraction vector; the third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is constructed repeatedly, the optimal solution is obtained by optimization, and the optimal solution is projected according to the exact solution subspace search method, until the iterative search obtains the third blind extraction vector. M Blind extraction vectors;
[0080] The source signal extraction module is used to use all the blind extraction vectors obtained above as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0081] For the specific definition of the signal blind extraction device without error accumulation, please refer to the definition of the signal blind extraction method without error accumulation above, which will not be repeated here. Each module in the above-mentioned signal blind extraction device without error accumulation can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0082] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a signal blind extraction method without error accumulation is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.
[0083] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0084] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0085] Get the observed signal vector; where the observed signal vector contains the M Source signal;
[0086] Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix;
[0087] According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector;
[0088] The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is reconstructed based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching;
[0089] The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors;
[0090] All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0091] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0092] Get the observed signal vector; where the observed signal vector contains the M Source signal;
[0093] Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix;
[0094] According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector;
[0095] The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observed signal vector is reconstructed based on the first blind extraction vector and the second blind extraction vector and the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching;
[0096] The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors;
[0097] All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and based on the blind extraction matrix M The row vectors are hidden in the observed signal vector M The source signal is extracted to obtain the source signal extraction result; wherein, based on the The estimated error of the source signal extraction result obtained by the row vector is not affected by the The influence of the estimation error of the source signal extraction result obtained by the row vector; among them, .
[0098] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink), DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0099] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0100] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A blind signal extraction method without error accumulation, characterized in that: The method comprises: Obtain an observation signal vector; wherein the observation signal vector contains the M Source signal; Estimate a first blind extraction vector according to the covariance matrix of the observed signal vector, and use the first blind extraction vector as the first row vector in the blind extraction matrix; According to the first blind extraction vector and based on the blind extraction criterion, a blind extraction cost function of the observed signal vector is constructed, a gradient descent optimization algorithm is used to search for an optimal solution of the blind extraction cost function, and according to an exact solution subspace search method, the optimal solution is projected into a subspace orthogonal to the first blind extraction vector, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector; The second blind extraction vector is used as the second row vector in the blind extraction matrix, and the blind extraction cost function of the observation signal vector is re-constructed based on the first blind extraction vector and the second blind extraction vector and based on the blind extraction criterion, and the gradient descent optimization algorithm is used again to search for the optimal solution of the blind extraction cost function, and according to the exact solution subspace search method, the optimal solution is again projected into a subspace orthogonal to both the first blind extraction vector and the second blind extraction vector, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and a third blind extraction vector is obtained by searching; The third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is repeatedly constructed, the optimal solution is obtained by optimization, and the optimal solution projection is performed according to the exact solution subspace search method until the iterative search obtains the third row vector in the blind extraction matrix. M Blind extraction vectors; All the blind extraction vectors obtained above are used as the blind extraction matrix M row vectors, and according to the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the p The estimated error of the source signal extraction result obtained by the row vector is not affected by the p -1 row vector to obtain the estimated error of the source signal extraction result; where, ; The method of estimating a first blind extraction vector according to the covariance matrix of the observed signal vector and using the first blind extraction vector as the first row vector in the blind extraction matrix includes: When the calculation time delay is 0, The observed signal vector at time The covariance matrix of ; According to the covariance matrix The matrix consisting of the left singular vectors of , further calculations yield the matrix and matrix ;in, Indicates the time delay is hour, The covariance matrix of Indicates the time delay is hour, The covariance matrix of , represents the number of covariance matrices; Among them, the matrix and matrix There are several generalized eigenvalues, and the generalized eigenvector corresponding to the largest generalized eigenvalue is used as the first blind extraction vector , and the first blind extraction vector As a blind extraction matrix The first row vector in ; Among them, the blind extraction cost function of the observed signal vector is constructed according to the blind extraction vector and based on the blind extraction criterion, which is expressed as ;in, is the blind extraction vector, the superscript T represents transpose, is the matrix and matrix Construct the intermediate matrix obtained, and The row vector of , Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors; , ; Wherein, the gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, including: Using gradient descent optimization algorithm to search blind extraction cost function The optimal solution ; Wherein, the gradient descent optimization algorithm is expressed as , ;in, Represents the blind extraction vector The Frobenius norm of , is the learning rate, Blind extraction vector The iterative update rule is: Represents the blind extraction vector It plays a normalizing role.
2. The method according to claim 1, characterized in that Get the observed signal vector, including: Get The observed signal vector at time t is expressed as ,in, No. Observed signal, superscript T represents transpose, L is the number of observed signals in the observed signal vector, and the observed signal vector contains M Source signals to be extracted .
3. The method according to claim 1, characterized in that The blind extraction criterion is expressed as: , and parallel.
4. The method according to claim 1, characterized in that The optimal solution projection is performed according to the exact solution subspace search method, including: For the previously acquired p Blind extraction vector Transpose to obtain the transposed blind extraction vector , and through Perform singular value decomposition, corresponding to the singular value 0 Left singular vectors The space formed is the same as the previous one p Blind extraction vector orthogonal subspace; where , M is the number of source signals to be extracted hidden in the observed signal vector; According to the exact solution subspace search method, the blind extraction cost function The optimal solution Project to the previous p Blind extraction vector In the orthogonal subspace, the optimal solution of the blind extraction cost function is not equal to the estimated prior p Any one of the blind extraction vectors; judge Whether the iterative process converges, if not, reconstruct the blind extraction cost function and solve the optimal solution; if converged, according to the subspace formed by all optimal solutions of the blind extraction cost function and does not contain the previous Blind extraction vector Subspace of The intersection of Blind extraction vector .
5. A signal blind extraction device without error accumulation, characterized in that: The device comprises: A data acquisition module is used to acquire an observation signal vector; wherein the observation signal vector contains the data to be extracted. M Source signal; A first blind extraction vector estimation module, used for estimating a first blind extraction vector according to a covariance matrix of the observed signal vector, and using the first blind extraction vector as a first row vector in a blind extraction matrix; The blind extraction vector iterative search module is used to construct a blind extraction cost function of the observed signal vector according to the first blind extraction vector and based on a blind extraction criterion, use a gradient descent optimization algorithm to search for an optimal solution of the blind extraction cost function, and project the optimal solution into a subspace orthogonal to the first blind extraction vector according to an exact solution subspace search method, thereby limiting the optimal solution to be not equal to the first blind extraction vector, and searching to obtain a second blind extraction vector; the second blind extraction vector is used as a second row vector in the blind extraction matrix, and the blind extraction matrix is constructed again according to the first blind extraction vector and the second blind extraction vector and based on the blind extraction criterion. The blind extraction cost function of the observed signal vector is constructed, and the optimal solution of the blind extraction cost function is searched again by the gradient descent optimization algorithm, and the optimal solution is projected again to a subspace orthogonal to the first blind extraction vector and the second blind extraction vector according to the exact solution subspace search method, so as to limit the optimal solution to be not equal to the first blind extraction vector and the second blind extraction vector, and the third blind extraction vector is obtained by searching; the third blind extraction vector is used as the third row vector in the blind extraction matrix, and the blind extraction cost function is constructed repeatedly, the optimal solution is obtained by optimizing, and the optimal solution is projected according to the exact solution subspace search method, until the iterative search obtains the first blind extraction vector. M Blind extraction vectors; The source signal extraction module is used to use all the blind extraction vectors obtained above as the blind extraction matrix M row vectors, and according to the blind extraction matrix M The row vectors represent the hidden M The source signal is extracted to obtain the source signal extraction result; wherein, based on the p The estimated error of the source signal extraction result obtained by the row vector is not affected by the p -1 row vector to obtain the estimated error of the source signal extraction result; where, ; The method of estimating a first blind extraction vector according to the covariance matrix of the observed signal vector and using the first blind extraction vector as the first row vector in the blind extraction matrix includes: When the calculation time delay is 0, The observed signal vector at time The covariance matrix of ; According to the covariance matrix The matrix consisting of the left singular vectors of , further calculations yield the matrix and matrix ;in, Indicates the time delay is hour, The covariance matrix of Indicates the time delay is hour, The covariance matrix of , represents the number of covariance matrices; Among them, the matrix and matrix There are several generalized eigenvalues, and the generalized eigenvector corresponding to the largest generalized eigenvalue is used as the first blind extraction vector , and the first blind extraction vector As a blind extraction matrix The first row vector in ; Among them, the blind extraction cost function of the observed signal vector is constructed according to the blind extraction vector and based on the blind extraction criterion, which is expressed as ;in, is the blind extraction vector, the superscript T represents transpose, is the matrix and matrix Construct the intermediate matrix obtained, and The row vector of is . ., Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors, Representation Matrix No. row vectors; , ; Wherein, the gradient descent optimization algorithm is used to search for the optimal solution of the blind extraction cost function, including: Using gradient descent optimization algorithm to search blind extraction cost function The optimal solution ; Wherein, the gradient descent optimization algorithm is expressed as , ;in, Represents the blind extraction vector The Frobenius norm of , is the learning rate, Blind extraction vector The iterative update rule is: Represents the blind extraction vector It plays a normalizing role.
6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.
