Method for intelligently estimating number of information sources of multiple time-frequency overlapped signals under non-Gaussian interference

Abstract

The embodiment of the invention relates to the technical field of electromagnetic space signal processing, in particular to an intelligent estimation method for the number of information sources of multiple time-frequency overlapping signals under non-Gaussian interference, and the method comprises the steps: carrying out the sparse decomposition of an obtained time-frequency overlapping signal into a plurality of matching signals through a matching pursuit algorithm, a nonlinear converter is adopted to weaken the influence of alpha stable noise on time-frequency overlapping signals, a plurality of matching signals and original signals are utilized to form virtual multi-channel signals, and a generalized correlation matrix is constructed; performing data preprocessing operation including eigenvalue decomposition and similarity transformation on the generalized correlation matrix to obtain a processed generalized correlation matrix; and inputting the processed generalized correlation matrix into an improved U-Net network to obtain a segmentation result output by the improved U-Net network, and realizing intelligent estimation of the number of time-frequency overlapped multiple signal sources under the non-Gaussian interference according to the segmentation result. The method is high in estimation precision, the estimation performance is not influenced by the modulation mode and the frequency spectrum aliasing degree, and the robustness is high.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of electromagnetic space signal processing, and in particular to a method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference. Background Art

[0002] Estimation of the number of information sources originates from the field of array signal processing technology, and is also an important research topic in the field of electromagnetic space signal processing technology. Only after obtaining the accurate number of information sources can subsequent work be carried out. Common methods for estimating the number of information sources are based on overdetermined or positively determined conditions, that is, the number of receiving channels is greater than or equal to the number of information sources. For example, estimation methods based on the Akaike information criterion, estimation methods based on the minimum description length criterion, algorithms based on Gale circle estimation, and singular value decomposition methods. Both the AIC criterion and the MDL criterion can be well applied to practical engineering, but under small sample conditions, the AIC criterion and the MDL criterion will have problems of overestimation and underestimation, respectively. At the same time, both the AIC criterion and the MDL criterion require the noise to be Gaussian white noise, and are only applicable to overdetermined or positively determined conditions. In underdetermined conditions, the number of information sources cannot be accurately estimated.

[0003] In recent years, domestic and foreign teams have conducted research on the problem of estimating the number of sources in underdetermined situations, especially in the special case of underdetermined single channels. Using other auxiliary methods to expand the dimension of the time-frequency overlapping signal, converting the time-frequency overlapping signal into a virtual multi-channel signal, and then using the multi-channel source number estimation method to estimate the number of sources on the virtual multi-channel signal is a common single-channel source number estimation idea. Pan et al. decomposed the time-frequency overlapping signal through empirical mode decomposition to obtain the intrinsic mode function, and combined it with the IMF to form a multi-dimensional signal to expand the dimension of the time-frequency overlapping signal. Then, the multi-channel source number estimation method was used to obtain the number of sources on the multi-dimensional signal. However, this type of method is greatly affected by the spectrum overlap rate and has certain limitations.

[0004] With the development of deep learning technology, deep learning technology has been gradually introduced into the research of source number estimation. Although it is in its infancy, some research results have been achieved. Zhang et al. directly sent the time signal as the input feature into the E2ECNN network to achieve end-to-end source number estimation. Andrei et al. designed a deep stacked convolutional network with a corresponding structure for the problem of estimating the number of mixed speech signals. However, the above method is for estimating the number of speech signals, and whether it is applicable to communication signals needs further verification. Shen et al. reconstructed the received complex signal into in-phase and orthogonal data to adapt CNN, and used the powerful feature extraction capability of CNN to extract the deep features of the mixed signal, thereby estimating the number of sources. At present, the proposed source number estimation methods based on deep learning are all in their infancy. The neural network models used are mostly classic shallow networks with relatively simple structures, and their performance cannot meet actual needs.

[0005] Through the above analysis, the current spectrum prediction method has the following defects.

[0006] First, the currently proposed methods for estimating the number of signal sources require certain prior information or have better estimation performance for specific mixed signals, which has certain limitations.

[0007] Second, most of the proposed methods for estimating the number of signal sources consider the influence of Gaussian noise, and there has been no in-depth research on the number estimation under the conditions where non-Gaussian interference and Gaussian noise coexist.

[0008] Third, although the performance of the source number estimation method based on deep learning has been improved compared with the traditional method, the models used are mostly shallow networks, and the extraction of input features is not sufficient. Summary of the invention

[0009] To solve the above technical problems, an embodiment of the present application provides an intelligent method for estimating the number of sources of time-frequency overlapping multiple signals under non-Gaussian interference, which effectively solves the problem of estimating the number of sources of time-frequency overlapping multiple signals under the coexistence of Gaussian noise and alpha-stable noise, and the overall estimation effect is superior. The estimation performance is not affected by the modulation mode and the degree of spectrum aliasing, and has good robustness.

[0010] To achieve the above-mentioned purpose, an embodiment of the present application proposes a method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference, and the method includes the following steps: sparsely decomposing the acquired time-frequency overlapping signal into multiple matching signals through a matching pursuit algorithm, and using a nonlinear transformer to weaken the influence of alpha stable noise on the time-frequency overlapping signal, using multiple matching signals and the original signal to form a virtual multi-channel signal, and constructing a generalized correlation matrix; performing data preprocessing operations on the generalized correlation matrix, increasing the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtaining a processed generalized correlation matrix; inputting the processed generalized correlation matrix into an improved U-Net network, finding the boundary between the signal eigenvalue and the noise eigenvalue, obtaining the segmentation result output by the improved U-Net network, and realizing intelligent estimation of the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference based on the segmentation result.

[0011] To achieve the above-mentioned purpose, an embodiment of the present application also proposes a system for intelligently estimating the number of sources of time-frequency overlapping multiple signals under non-Gaussian interference, the system comprising: a matrix construction module, used to sparsely decompose the acquired time-frequency overlapping signal into multiple matching signals through a matching pursuit algorithm, and use a nonlinear transformer to weaken the influence of alpha stable noise on the time-frequency overlapping signal, and use multiple matching signals and the original signal to form a virtual multi-channel signal to construct a generalized correlation matrix; a data preprocessing module, used to perform data preprocessing operations on the generalized correlation matrix, increase the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtain a processed generalized correlation matrix; a source number estimation module, used to input the processed generalized correlation matrix into an improved U-Net network, find the boundary between the signal eigenvalue and the noise eigenvalue, obtain the segmentation result output by the improved U-Net network, and realize intelligent estimation of the number of sources of time-frequency overlapping multiple signals under non-Gaussian interference based on the segmentation result.

[0012] To achieve the above-mentioned purpose, an embodiment of the present application also proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference as described above.

[0013] To achieve the above-mentioned purpose, an embodiment of the present application also proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference as described above.

[0014] The embodiment of the present application proposes an intelligent estimation method for the number of sources of time-frequency overlapping multi-signals under non-Gaussian interference. The time-frequency overlapping signals are sparsely decomposed into multiple matching signals by a matching pursuit algorithm, and the influence of alpha stable noise on the time-frequency overlapping signals is weakened by a nonlinear transformer, so as to construct a generalized correlation matrix, perform eigenvalue decomposition and similarity transformation operations on the generalized correlation matrix, increase the difference between the signal eigenvalue and the noise eigenvalue, and obtain the processed generalized correlation matrix. The processed generalized correlation matrix is ​​regarded as a "two-dimensional image", each signal eigenvalue on the processed generalized correlation matrix is ​​regarded as an independent source target, and the noise eigenvalue is regarded as the background of the "two-dimensional image". Finally, the semantic segmentation between the source target and the noise background is completed by an improved U-Net network, and the number of sources can be obtained according to the output segmentation result. The problem of estimating the number of sources of time-frequency overlapping multi-signals under the coexistence of Gaussian noise and alpha stable noise is effectively solved, and the overall estimation effect is superior, and the estimation performance is not affected by the modulation mode and the spectrum aliasing degree, and has good robustness.

[0015] Optionally, before sparsely decomposing the acquired time-frequency overlapping signal into a plurality of matching signals by a matching pursuit algorithm, the method further includes:

[0016] Obtain the time-frequency overlapping signal and establish a time-frequency overlapping signal receiving model. The signal components in the time-frequency overlapping signal are independent of each other, and the signal components and noise are also independent of each other. The time-frequency overlapping signal receiving model is expressed by the formula:

[0017]

[0018] Where M is the total number of signal components, s i (t) represents the i-th signal component, h i is a preset constant, w(t) represents Gaussian white noise, u(t) represents alpha stable distribution noise, A i For i (t) signal amplitude, a i (c) represents the transmitted codeword sequence, C is the total number of transmitted codewords, and f i For i (t) is the carrier frequency, is the initial phase, p i (t) is the shaped pulse function with a roll-off factor of α = 0.5, T si is the symbol period, and x(t) represents the time-frequency overlapping signal reception model.

[0019] Optionally, a symmetric alpha stable distribution is used in the time-frequency overlapping signal reception model to characterize non-Gaussian interference, and the characteristic function expression of non-Gaussian interference is:

[0020] φ(w)=exp(jμw-γ|w| ζ );

[0021] Among them, φ(w) is the characteristic function of non-Gaussian interference, ζ is the characteristic factor, which is used to control the thickness of the tail in the alpha stable distribution, indicating the strength of the pulse, μ is the position parameter, which is used to control the center position of the alpha stable distribution, indicating the offset of the probability density function on the horizontal axis, and γ is the scale parameter, which is used to measure the degree to which the alpha stable distribution deviates from the mean.

[0022] Optionally, the acquired time-frequency overlapping signal is sparsely decomposed into multiple matching signals by a matching pursuit algorithm, and a nonlinear transformer is used to reduce the influence of alpha stable noise on the time-frequency overlapping signal, and multiple matching signals and the original signal are used to form a virtual multi-channel signal, including:

[0023] Based on the time-frequency overlapping signal receiving model, the discrete expression of the single-channel receiving signal is established. The discrete expression of the single-channel receiving signal is:

[0024]

[0025] The matching pursuit algorithm is used to sparsely represent x(n). The sparse representation of x(n) is:

[0026]

[0027] Among them, K is the total number of iterations, r k (n) is the residual signal after the kth iteration, g k (n) is the same as r k (n) Gabor atom with the largest inner product, <·,·> indicates inner product operation;

[0028] Save the projection of the best matching atom in each iteration, that is, save <r k (n),g k (n)>·g k (n), and use it as a dimension in the virtual multi-channel signal, and finally obtain multiple matching signals, which are expressed by the formula:

[0029]

[0030] Optionally, a generalized correlation matrix is ​​constructed using the following formula:

[0031]

[0032] Among them, f[y i (n)] represents the nonlinear transformation to suppress the interference of alpha noise. If |yi (n)|<λ i , then f[y i (n)]=1, if |y i (n)|≥λ i , then f[y i (n)]=|y i (n)|+|ξ|, |ξ| is the compression factor, for |y i The median of (n)|, G ij Represents the element in the i-th row and j-th column of G.

[0033] Optionally, a data preprocessing operation is performed on the generalized correlation matrix, and the difference between the signal eigenvalue and the noise eigenvalue is increased by an eigenvalue decomposition operation and a similarity transformation operation to obtain a processed generalized correlation matrix, including:

[0034] The generalized correlation matrix G is divided into blocks as follows:

[0035]

[0036] Among them, G 1 is a square matrix consisting of the elements of the first K-1 rows and columns of G, and g is a column vector consisting of the elements of the first K-1 rows of the Kth column of G. H is a row vector consisting of the elements of the first K-1 columns of the Kth row of G;

[0037] To the opponent G 1 Perform eigenvalue decomposition, which is implemented by the following formula:

[0038]

[0039] T 1 =[t 1 ,t 2 ,…,t K-1 ];

[0040] ∑ 1 =diag(τ 1 ,τ 2 ,…,τ K-1 );

[0041] Among them, T 1 By G 1 The unitary matrix composed of the eigenvectors of K-1 Represents G 1 The K-1th eigenvector of 1 It's G 1 The diagonal matrix of the eigenvalues ​​of K-1 Represents G1 The K-1th eigenvalue of G is the same as G 1 The eigenvalues ​​of satisfy the following relationship:

[0042] λ 1 ≥τ 1 …≥λ M+1 ≥τ M+1 ≥…λ K-1 ≥τ K-1 ≥λ K ;

[0043] Among them, λ K represents the Kth eigenvalue of G;

[0044] make Based on T 2 Perform unitary transformation on G to obtain the unitary transformation matrix G 2 , unitary transformation matrix G 2 It is expressed by the formula:

[0045]

[0046] In ∑ 1 Based on this, a diagonal matrix P is constructed. The diagonal matrix P is expressed by the formula:

[0047] P = diag (p 1 ,p 2 ,…,p K );

[0048]

[0049] k=1,2,…,K-1;

[0050]

[0051] Use the diagonal matrix P to transform the unitary matrix G 2 Perform similarity transformation and obtain the similarity transformation matrix G 3 , similarity transformation matrix G 3 That is, the processed generalized correlation matrix, the similarity transformation matrix G 3 It is expressed by the formula:

[0052]

[0053] Optionally, the processed generalized correlation matrix is ​​input into an improved U-Net network, the boundary between the signal eigenvalue and the noise eigenvalue is found, and the segmentation result of the improved U-Net network output is obtained. According to the segmentation result, the number of information sources of time-frequency overlapping multi-signals under non-Gaussian interference is intelligently estimated, including: reducing the four-layer structure of the U-Net network to two layers, adding a batch normalization layer and a ReLU activation function layer after each convolution layer of the original encoding operation in the encoder of the U-Net network, adding a shortcut connection outside the convolution layer to add the output of the convolution layer to the original input, thereby obtaining an improved U-Net network; the similarity transformation matrix G 3 Input into the improved U-Net network and output the similarity transformation matrix G 3 The normalized probability that each pixel point in belongs to signal or noise is determined, the dividing line between the signal eigenvalue and the noise eigenvalue is determined, and the segmentation result output by the improved U-Net network is obtained; each row in the segmentation result is traversed in turn, all elements of the current row are accumulated to obtain a K-dimensional vector, and the number of elements in the K-dimensional vector whose values ​​are greater than K / 2 is counted to obtain the number of information sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related technologies, the drawings required for use in the embodiments of the present application or the related technical descriptions will be briefly introduced below. Obviously, the following drawings are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The drawings described here are only used to explain the present application and are not used to limit the present application.

[0055] Figure 1 It is a flow chart of a method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference provided in one embodiment of the present application;

[0056] Figure 2 It is a performance diagram of a method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference provided in an embodiment of the present application;

[0057] Figure 3 It is a structural schematic diagram of a system for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference provided in another embodiment of the present application;

[0058] Figure 4 It is a structural schematic diagram of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. In the various embodiments of the present application, many technical details are proposed in order to make the reader better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The division of the following embodiments is only for the convenience of description, and the specific implementation mode of the present application should not constitute any limitation. The various embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0060] An embodiment of the present application proposes a method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference, which is applied to an electronic device, wherein the electronic device may be a terminal or a server. The electronic device in this embodiment and the following embodiments is described using a server as an example. The implementation details of the method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this embodiment are described in detail below. The following content is only the implementation details provided for ease of understanding and is not necessary for implementing this solution.

[0061] The specific process of the method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this embodiment can be as follows: Figure 1 As shown, including:

[0062] S1, the obtained time-frequency overlapping signal is sparsely decomposed into multiple matching signals through the matching pursuit algorithm, and the nonlinear transformer is used to weaken the influence of alpha stable noise on the time-frequency overlapping signal. Multiple matching signals and the original signal are used to form a virtual multi-channel signal to construct a generalized correlation matrix.

[0063] In the specific implementation, the server first needs to obtain the time-frequency overlapping signal, and then sparsely decompose the acquired time-frequency overlapping signal into multiple matching signals through the matching pursuit algorithm, and use a nonlinear transformer to weaken the influence of alpha stable noise on the time-frequency overlapping signal. Then, multiple matching signals are used together with the original signal to form a virtual multi-channel signal to construct a generalized correlation matrix.

[0064] In one example, the server needs to acquire the time-frequency overlapping signal in real time and establish a time-frequency overlapping signal receiving model, in which each signal component is independent of each other, and each signal component is also independent of noise.

[0065] In one example, the modulation method of the transmitted signal may be MPSK or 16QAM.

[0066] In one example, the time-frequency overlapping signal reception model can be expressed by the formula:

[0067]

[0068] Where M is the total number of signal components, s i (t) represents the i-th signal component, h i is a preset constant, w(t) represents Gaussian white noise, u(t) represents alpha stable distribution noise, A i For i (t) signal amplitude, a i (c) represents the transmitted codeword sequence, C is the total number of transmitted codewords, and f i For i (t) is the carrier frequency, is the initial phase, p i (t) is the shaped pulse function with a roll-off factor of α = 0.5, t si is the symbol period, and x(t) represents the time-frequency overlapping signal reception model.

[0069] In one example, a symmetric alpha stable distribution is used in a time-frequency overlapping signal reception model to characterize non-Gaussian interference. The characteristic function expression of non-Gaussian interference is:

[0070] φ(w)=exp(jμw-γ|w| ζ );

[0071] Among them, φ(w) is the characteristic function of non-Gaussian interference, ζ is the characteristic factor, which is used to control the thickness of the tail in the alpha stable distribution, indicating the strength of the pulse, μ is the position parameter, which is used to control the center position of the alpha stable distribution, indicating the offset of the probability density function on the horizontal axis, and γ is the scale parameter, which is used to measure the degree to which the alpha stable distribution deviates from the mean.

[0072] In one example, the server first needs to establish a discrete expression of a single-channel received signal based on a time-frequency overlapping signal reception model. The discrete expression of a single-channel received signal is:

[0073]

[0074] Then, the matching pursuit algorithm is used to sparsely represent x(n), and the sparse representation of x(n) is:

[0075]

[0076] Among them, K is the total number of iterations, r k (n) is the residual signal after the kth iteration, g k (n) is the same as r k (n) The Gabor atom with the largest inner product, <·,·> indicates an inner product operation, n = 1, 2, …, N, where N is the total number of sampling points.

[0077] Finally, the server saves the projection of the best matching atom in each iteration, that is, saves the sparse representation of x(n) <r k (n),g k (n)>·g k (n), and use it as a dimension in the virtual multi-channel signal, and finally obtain multiple matching signals, which are expressed by the formula:

[0078]

[0079] For the convenience of representation, let y k+1 (n)= <r k (n),g k (n)>·g k (n).

[0080] In an example, the server constructs a generalized correlation matrix G based on Y. The generalized correlation matrix G can be expressed by the formula:

[0081]

[0082] Among them, f[y i (n)] represents the nonlinear transformation to suppress the interference of alpha noise. If |y i (n)|<λ i , then f[y i (n)]=1, if |y i (n)|≥λ i , then f[y i (n)]=|y i (n)|+|ξ|, |ξ| is the compression factor, for |y i The median of (n)|, G ij Represents the element in the i-th row and j-th column of G.

[0083] S2, performing data preprocessing operations on the generalized correlation matrix, increasing the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtaining the processed generalized correlation matrix.

[0084] In a specific implementation, after obtaining the generalized correlation matrix G, the server needs to perform preprocessing operations on the generalized correlation matrix G including eigenvalue decomposition operations and similarity transformation operations to increase the difference between the signal eigenvalues ​​and the noise eigenvalues ​​to obtain the processed generalized correlation matrix.

[0085] In one example, the server performs a preprocessing operation on the generalized correlation matrix G. First, the generalized correlation matrix G needs to be divided into blocks as follows:

[0086]

[0087] Among them, G 1 is a square matrix consisting of the elements of the first K-1 rows and columns of G, and g is a column vector consisting of the elements of the first K-1 rows of the Kth column of G. H is a row vector consisting of the elements of the first K-1 columns of the Kth row of G.

[0088] Next, the server will 1 Perform eigenvalue decomposition, which is implemented by the following formula:

[0089]

[0090] T 1 =[t 1 ,t 2 ,…,t K-1 ];

[0091] ∑ 1 =diag(τ 1 ,τ 2 ,…,τ K-1 );

[0092] Among them, T 1 By G 1 The unitary matrix composed of the eigenvectors of K-1 Represents G 1 The K-1th eigenvector of 1 It's G 1 The diagonal matrix of the eigenvalues ​​of K-1 Represents G 1 The K-1th eigenvalue of G is the same as G 1 The eigenvalues ​​of satisfy the following relationship:

[0093] λ 1 ≥τ 1 …≥λ M+1 ≥τ M+1 ≥…λ K-1 ≥τ K-1 ≥λ K ;

[0094] Among them, λ K represents the Kth eigenvalue of G.

[0095] The server then needs to Based on T 2Perform unitary transformation on G to obtain the unitary transformation matrix G 2 , unitary transformation matrix G 2 It is expressed by the formula:

[0096]

[0097] After completing the unitary transformation, we get the unitary transformation matrix G 2 Afterwards, the server 1 Based on this, a diagonal matrix P is constructed. The diagonal matrix P is expressed by the formula:

[0098] P = diag (p 1 ,p 2 ,…,p K );

[0099]

[0100] k=1,2,…,K-1;

[0101]

[0102] Finally, the server uses the diagonal matrix P to transform the unitary matrix G 2 Perform similarity transformation and obtain the similarity transformation matrix G 3 , similarity transformation matrix G 3 That is, the processed generalized correlation matrix, the similarity transformation matrix G 3 It is expressed by the formula:

[0103]

[0104] Similarity transformation matrix G 3 With the unitary transformation matrix G 2 are similar matrices, they have the same eigenvalues ​​and are proportional to each other by different factors p. 1 / p K The original values ​​are compressed, so that the boundary between the signal characteristic value and the noise characteristic value can be more obvious, which is more conducive to distinguishing between signals and noise.

[0105] S3, input the processed generalized correlation matrix into the improved U-Net network, find the boundary between the signal eigenvalue and the noise eigenvalue, obtain the segmentation result output by the improved U-Net network, and realize the intelligent estimation of the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference based on the segmentation result.

[0106] In one example, after completing the data preprocessing operation, the server obtains the processed generalized correlation matrix G (i.e., the similarity transformation matrix G 3), the processed generalized correlation matrix G can be input into the improved U-Net network to find the boundary between the signal eigenvalue and the noise eigenvalue, and the segmentation result output by the improved U-Net network can be obtained. According to the segmentation result, the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference can be intelligently estimated.

[0107] In one example, in order to make the U-Net network have a similarity transformation matrix G that can handle small dimensions 3 To improve the capacity of the server, the U-Net network needs to be improved, reducing the four-layer structure of the U-Net network to two layers, adding a batch normalization layer and a ReLU activation function layer after each convolution layer of the original encoding operation in the encoder of the U-Net network, and adding a shortcut connection outside the convolution layer to add the output of the convolution layer to the original input, thereby obtaining an improved U-Net network. Such an improved U-Net network helps to retain the information of the original input and avoid losing important features.

[0108] Next, the server converts the similarity transformation matrix G 3 Input into the improved U-Net network and output the similarity transformation matrix G 3 The normalized probability that each pixel belongs to signal or noise is determined, the dividing line between the signal eigenvalue and the noise eigenvalue is determined, and the segmentation result output by the improved U-Net network is obtained.

[0109] Finally, the server traverses each row in the segmentation result in turn, accumulates all elements of the current row to obtain a K-dimensional vector, and counts the number of elements in the K-dimensional vector whose values ​​are greater than K / 2. The number of elements in the K-dimensional vector whose values ​​are greater than K / 2 is the number of information sources.

[0110] In this embodiment, the time-frequency overlapping signal is sparsely decomposed by a matching pursuit algorithm, and a nonlinear transformer is used to reduce the influence of alpha stable noise on the time-frequency overlapping signal, so as to construct a generalized correlation matrix, perform eigenvalue decomposition operation and similarity transformation operation on the generalized correlation matrix, increase the difference between the signal eigenvalue and the noise eigenvalue, and obtain the processed generalized correlation matrix. The processed generalized correlation matrix is ​​regarded as a "two-dimensional image", each signal eigenvalue on the processed generalized correlation matrix is ​​regarded as an independent source target, and the noise eigenvalue is regarded as the background of the "two-dimensional image". Finally, the semantic segmentation between the source target and the noise background is completed by an improved U-Net network, and the number of sources can be obtained according to the output segmentation result. The problem of estimating the number of sources of time-frequency overlapping multi-signals under the coexistence of Gaussian noise and alpha stable noise is effectively solved, and the overall estimation effect is superior, and the estimation performance is not affected by the modulation mode and the spectrum aliasing degree, and has good robustness.

[0111] The step division of the above methods is only for clear description. They can be combined into one step or some steps can be decomposed into multiple steps during implementation. As long as they include the same logical relationship, they are all within the scope of protection of this application. Adding insignificant modifications or introducing insignificant designs to the algorithm or process without changing the core design of the algorithm and process are all within the scope of protection of this application.

[0112] In an example, in order to evaluate the intelligent estimation method for the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this application, we conducted relevant simulation experiments. Through simulation experiments, we verified the estimation performance of the intelligent estimation method for the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this application under the coexistence of Gaussian noise and alpha-stable noise. Specifically, the correct estimation probability under different signal-to-noise ratios was used as the performance evaluation criterion.

[0113] The parameters in the simulation experiment are set as follows: the modulation types of the independent signal sources in the time-frequency overlapping signal are BPSK, QPSK, 8PSK and 16QAM, and the symbol rate is f b =1000Baud, the carrier frequencies are set to f c0 =3800Hz, f c1 =4400Hz, f c2 =5000Hz, f c3 =5600Hz, the spectrum aliasing degree is 68.4%, the channel is a Gaussian noise channel, and the sampling rate is f s =16000Hz, the length of each sample data is 12800, and the signal components are mixed with equal power.

[0114] The signal-to-noise ratio range is set to -7dB to 7dB. 1600 sample data are generated under each signal-to-noise ratio, including 400 samples of one, two, three, and four mixed signal sources, generating a total of 24,000 sample data sets. The network model is trained in the Pytorch1.11.0 environment. The concepts of signal-to-noise ratio (SNR) and signal-to-interference ratio (SIR) are introduced, which represent the power ratio of the signal to Gaussian noise and the power ratio of the signal to alpha stable distribution noise interference, respectively.

[0115] Figure 2The relationship between the probability of correct estimation and the number of information sources under different signal-to-noise ratios is shown, which can illustrate that the intelligent estimation method for the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this application can effectively estimate the number of information sources under the coexistence of Gaussian noise and alpha stable noise. In addition, under the same signal-to-noise ratio conditions, the probability of correct estimation gradually decreases with the increase in the number of information sources. This is because the more types of signals in the mixed signal under the same signal-to-noise ratio, the lower the equivalent signal-to-noise ratio for each of the signals, which leads to a decrease in the estimation performance of the intelligent estimation method for the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this application.

[0116] Another embodiment of the present application proposes an intelligent estimation system for the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference. The implementation details of the intelligent estimation system for the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this embodiment are described in detail below. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this example.

[0117] Figure 3 It is a structural diagram of a system for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference proposed in this embodiment, and the system includes: a matrix construction module M1, a data preprocessing module M2 and a information source number estimation module M3.

[0118] The matrix construction module M1 is used to sparsely decompose the acquired time-frequency overlapping signal into multiple matching signals through the matching pursuit algorithm, and use a nonlinear transformer to weaken the influence of alpha stable noise on the time-frequency overlapping signal, and use multiple matching signals and the original signal to form a virtual multi-channel signal to construct a generalized correlation matrix.

[0119] The data preprocessing module M2 is used to perform data preprocessing operations on the generalized correlation matrix, increase the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtain the processed generalized correlation matrix.

[0120] The signal source number estimation module M3 is used to input the processed generalized correlation matrix into the improved U-Net network, find the boundary between the signal eigenvalue and the noise eigenvalue, obtain the segmentation result output by the improved U-Net network, and realize the intelligent estimation of the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference based on the segmentation result.

[0121] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application, but this does not mean that there are no other units in this embodiment.

[0122] It is not difficult to find that this embodiment is a system embodiment corresponding to the above-mentioned method embodiments, and this embodiment can be implemented in conjunction with the above-mentioned method embodiments. The relevant technical details and technical effects mentioned in the above-mentioned method embodiments are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned method embodiments.

[0123] Another embodiment of the present application provides an electronic device, wherein the specific structure of the electronic device is as follows: Figure 4 As shown, it includes: at least one processor C1; and a memory C2 that is communicatively connected to the at least one processor C1; wherein the memory C2 stores instructions that can be executed by the at least one processor C1, and the instructions are executed by the at least one processor C1 so that the at least one processor C1 can execute a method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference as described in the above-mentioned method embodiments.

[0124] The memory and the processor may be connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be further described herein. The bus interface is responsible for providing an interface between the bus and the transceiver. The transceiver may be one component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium.

[0125] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0126] Another embodiment of the present application proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference as described in the above method embodiments.

[0127] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned method embodiments can be implemented by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions for making a device (such as a single-chip microcomputer, chip, etc.) or a processor (processor) perform all or part of the steps in the above-mentioned method embodiments. The storage medium includes: a U disk, a mobile hard disk, a read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, and other media that can store program codes.

[0128] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A method for intelligently estimating the number of information sources of time-frequency overlapping multi-signals under non-Gaussian interference, characterized in that: The method comprises: The obtained time-frequency overlapping signal is sparsely decomposed into multiple matching signals through the matching pursuit algorithm, and the influence of alpha stable noise on the time-frequency overlapping signal is weakened by using nonlinear transformer. Multiple matching signals and the original signal are used to form a virtual multi-channel signal to construct a generalized correlation matrix. Performing data preprocessing operations on the generalized correlation matrix, increasing the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtaining the processed generalized correlation matrix; The processed generalized correlation matrix is ​​input into the improved U-Net network to find the boundary between the signal eigenvalue and the noise eigenvalue, and the segmentation result output by the improved U-Net network is obtained. Based on the segmentation result, the intelligent estimation of the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference is realized.

2. According to claim 1, the method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference is characterized in that: Before sparsely decomposing the acquired time-frequency overlapping signal into a plurality of matching signals by a matching pursuit algorithm, the method further includes: Obtain the time-frequency overlapping signal and establish a time-frequency overlapping signal receiving model. The signal components in the time-frequency overlapping signal are independent of each other, and the signal components and noise are also independent of each other. The time-frequency overlapping signal receiving model is expressed by the formula: Where M is the total number of signal components, s i (t) represents the i-th signal component, h i is a preset constant, w(t) represents Gaussian white noise, u(t) represents alpha stable distribution noise, A i For i (t) signal amplitude, a i (c) represents the transmitted codeword sequence, C is the total number of transmitted codewords, and f i For i (t) is the carrier frequency, is the initial phase, p i (t) is the shaped pulse function with a roll-off factor of α = 0.5, T si is the symbol period, and x(t) represents the time-frequency overlapping signal reception model.

3. The method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference according to claim 2, characterized in that: In the time-frequency overlapping signal receiving model, the symmetric alpha stable distribution is used to characterize non-Gaussian interference. The characteristic function expression of non-Gaussian interference is: φ(w)=exp(jµw-γ|w| ζ ); Among them, φ(w) is the characteristic function of non-Gaussian interference, ζ is the characteristic factor, which is used to control the thickness of the tail in the alpha stable distribution, indicating the strength of the pulse, μ is the position parameter, which is used to control the center position of the alpha stable distribution, indicating the offset of the probability density function on the horizontal axis, and γ is the scale parameter, which is used to measure the degree to which the alpha stable distribution deviates from the mean.

4. The method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference according to claim 3 is characterized in that: The obtained time-frequency overlapping signal is sparsely decomposed into multiple matching signals through the matching pursuit algorithm, and the influence of alpha stable noise on the time-frequency overlapping signal is weakened by a nonlinear transformer. Multiple matching signals and the original signal are used to form a virtual multi-channel signal, including: Based on the time-frequency overlapping signal receiving model, the discrete expression of the single-channel receiving signal is established. The discrete expression of the single-channel receiving signal is: The matching pursuit algorithm is used to sparsely represent x(n). The sparse representation of x(n) is: Among them, K is the total number of iterations, r k (n) is the residual signal after the kth iteration, g k (n) is the same as r k (n) Gabor atom with the largest inner product, <·, ·> indicates inner product operation; Save the projection of the best matching atom in each iteration, that is, save <r k (n), g k (n)>·g k (n), and use it as a dimension in the virtual multi-channel signal, and finally obtain multiple matching signals, which are expressed by the formula:

5. The method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference according to claim 4, characterized in that: The generalized correlation matrix is ​​constructed and implemented by the following formula: Among them, f[y i (n)] represents the nonlinear transformation to suppress the interference of alpha noise. If |y i (n)|<λ i , then f[y i (n)]=1, if |y i (n)|≥λ i , then f[y i (n)]=|y i (n)|+|ξ|, |ξ| is the compression factor, for |y i The median of (n)|, G ij Represents the element in row i and column j in G.

6. The method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference according to claim 5, characterized in that: The generalized correlation matrix is ​​subjected to data preprocessing operations, and the difference between the signal eigenvalue and the noise eigenvalue is increased through eigenvalue decomposition operations and similarity transformation operations, and the processed generalized correlation matrix is ​​obtained, including: The generalized correlation matrix G is divided into blocks as follows: Among them, G1 is a square matrix consisting of the elements of the first K-1 rows and the first K-1 columns of G, g is a column vector consisting of the elements of the first K-1 rows of the Kth column of G, and g H is a row vector consisting of the elements of the first K-1 columns of the Kth row of G; Perform eigenvalue decomposition on the square matrix G1. The eigenvalue decomposition is achieved by the following formula: T1=[t1,t2,...,t K-1 ]; ∑1=diag(τ1,τ2,…,τ K-1 ); Among them, T1 is the unitary matrix composed of the eigenvectors of G1, t K-1 represents the K-1th eigenvector of G1, ∑1 is the diagonal matrix of the eigenvalues ​​of G1, τ K-1 represents the K-1th eigenvalue of G1. The eigenvalue of G satisfies the following relationship with the eigenvalue of G1: λ1≥τ1…≥λ M+1 ≥τ M+1 ≥…λ K-1 ≥τ K-1 ≥λ K ; Among them, λ K represents the Kth eigenvalue of G; make Based on T2, a unitary transformation matrix G2 is obtained by performing a unitary transformation on G. The unitary transformation matrix G2 is expressed by the formula: The diagonal matrix P is constructed on the basis of ∑1. The diagonal matrix P is expressed by the formula: P=diag(p1,p2,…,p K ); k=1, 2, ..., K-1; The diagonal matrix P is used to perform a similarity transformation on the unitary transformation matrix G2 to obtain a similarity transformation matrix G3. The similarity transformation matrix G3 is the processed generalized correlation matrix. The similarity transformation matrix G3 is expressed by the formula:

7. The method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference according to claim 6, characterized in that: The processed generalized correlation matrix is ​​input into the improved U-Net network to find the boundary between the signal eigenvalue and the noise eigenvalue, and the segmentation result output by the improved U-Net network is obtained. Based on the segmentation result, the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference is intelligently estimated, including: The four-layer structure of the U-Net network is reduced to two layers, and a batch normalization layer and a ReLU activation function layer are added after each convolution layer of the original encoding operation in the encoder of the U-Net network. A shortcut connection is added outside the convolution layer to add the output of the convolution layer to the original input, thereby obtaining an improved U-Net network; The similarity transformation matrix G3 is input into the improved U-Net network, the normalized probability of each pixel in the similarity transformation matrix G3 belonging to the signal or the noise is output, the dividing line between the signal eigenvalue and the noise eigenvalue is determined, and the segmentation result output by the improved U-Net network is obtained; Traverse each row in the segmentation result in turn, accumulate all elements of the current row to obtain a K-dimensional vector, count the number of elements in the K-dimensional vector whose values ​​are greater than K / 2, and obtain the number of information sources.

8. An intelligent estimation system for the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference, characterized in that: The system comprises: The matrix construction module is used to sparsely decompose the acquired time-frequency overlapping signal into multiple matching signals through the matching pursuit algorithm, and use a nonlinear transformer to weaken the influence of alpha stable noise on the time-frequency overlapping signal, and use multiple matching signals and the original signal to form a virtual multi-channel signal to construct a generalized correlation matrix; A data preprocessing module is used to perform data preprocessing operations on the generalized correlation matrix, increase the difference between the signal eigenvalue and the noise eigenvalue through eigenvalue decomposition operations and similarity transformation operations, and obtain a processed generalized correlation matrix; The signal source number estimation module is used to input the processed generalized correlation matrix into the improved U-Net network, find the boundary between the signal eigenvalue and the noise eigenvalue, obtain the segmentation result output by the improved U-Net network, and realize the intelligent estimation of the number of signal sources of time-frequency overlapping multi-signals under non-Gaussian interference based on the segmentation result.

9. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a method for intelligently estimating the number of signal sources of time-frequency overlapping multiple signals under non-Gaussian interference as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement a method for intelligently estimating the number of information sources of time-frequency overlapping multiple signals under non-Gaussian interference as described in any one of claims 1 to 7.

Images