Communication signal anti-interference method based on shape parameter estimation and semi-blind source separation

Abstract

This invention discloses a communication signal anti-interference method based on shape parameter estimation and semi-blind source separation. The method includes the following steps: S1. A signal receiving device acquires a received signal through an antenna. The received signal is a mixed signal including a transmitted signal from a signal transmitting device and interference signals from other sources; S2. A signal shape parameter estimate of the received signal is obtained by combining deep learning and numerical analysis methods; S3. The received signal, the interference signal, and the signal shape parameter estimate are input into a signal processing system. The signal processing system uses a signal shape parameter estimation algorithm and an AuxIVA-based semi-blind source separation algorithm to acquire the target communication signal. This method can improve the accuracy of the algorithm, thereby effectively improving the anti-interference capability against co-band interference signals that cause interference in wireless communication.

Description

Technical Field

[0001] This invention relates to the fields of wireless communication technology and signal processing technology, specifically to a communication signal anti-interference method based on shape parameter estimation and semi-blind source separation. Background Technology

[0002] With the maturation of wireless communication technology and wireless networks, the modern electromagnetic environment has become highly complex and dense, filled with communication signals of various frequency bands, modulation schemes, and multiple access methods. Spectrum resource depletion and electromagnetic pollution are becoming increasingly serious problems. High-density spectrum occupancy and spectrum overlap have led to frequent interference between signals in adjacent frequency bands. As a crucial means of improving the reliability of communication systems, interference signal suppression and elimination have always played a vital role in communication system signal processing.

[0003] For communication signals affected by in-band interference, it is difficult to directly suppress and eliminate interference using frequency domain filtering. In such cases, blind source separation can be used to separate and eliminate the interference signal and extract and enhance the target signal. The most commonly used type of blind source separation algorithm is Independent Component Analysis (ICA). These algorithms require that the target signal and the interference signal are statistically independent, and that at most one source signal in the mixed signal follows a Gaussian distribution. A problem with ICA algorithms is that they can easily lead to permutation ambiguity. To address this issue, the academic community developed Independent Vector Analysis (IVA) algorithms, and to further accelerate the convergence speed, Auxiliary function-based Independent Vector Analysis (AuxIVA) algorithms emerged.

[0004] The main drawback of existing semi-blind source separation algorithms is their failure to accurately estimate the signal shape parameters, resulting in insufficient signal separation accuracy and weak anti-interference capability. This algorithm addresses this deficiency by improving the signal shape parameter estimation step, thereby enhancing signal separation accuracy and anti-interference capability. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention aims to provide a communication signal anti-interference method based on shape parameter estimation and semi-blind source separation. Compared with the prior art, this method can improve the accuracy of the algorithm and thus effectively improve the anti-interference capability against co-band interference signals that cause interference in wireless communication.

[0006] To achieve the above objectives, this invention provides a communication signal anti-interference method based on shape parameter estimation and semi-blind source separation, the method comprising the following steps:

[0007] S1. The signal receiving device acquires a received signal through an antenna, the received signal being a mixed signal including a transmitted signal from the signal transmitting device and interference signals from other sources;

[0008] S2. Combine deep learning and numerical analysis methods to obtain signal shape parameters. Estimated value;

[0009] S3. Combine the received signal, interference signal, and signal shape parameters. The estimated values ​​are input into the signal processing system, which uses a signal shape parameter estimation algorithm and an AuxIVA-based semi-blind source separation algorithm to obtain the target communication signal.

[0010] Furthermore, step S2 specifically includes the following steps:

[0011] S2.1 uses deep learning to obtain the estimated values ​​of the shape parameters of the first signal;

[0012] S2.2 The estimated values ​​of the shape parameters of the second signal are obtained by numerical analysis;

[0013] S2.3 The average of the first signal shape parameter estimate and the second signal shape parameter estimate is taken as the final result of the signal parameter estimate.

[0014] Furthermore, step S2.1 is implemented as follows: a neural network for signal shape parameter estimation is established. The signal shape parameter estimation is based on the time-frequency matrix of the signal. The overall network architecture includes two convolutional neural network layers, two average pooling layers, a flattening layer, and a fully connected layer.

[0015] Furthermore, the kernel size of both the CNN layer and the pooling layer is 3×3.

[0016] Furthermore, before the neural network algorithm used for signal shape parameter estimation is run, a set of random signals with specific shape parameter values ​​is first generated for network training. After training, the network is then used for signal shape parameter estimation.

[0017] Furthermore, step S2.2 is implemented as follows: Also based on the time-frequency representation matrix of the signal, the algorithm flow includes the following steps:

[0018] S2.2.1 First, the kurtosis value of the complex signal is calculated using the time-frequency matrix;

[0019] S2.2.2 Then, based on the CSK value and signal shape parameters... The relationship between values, for The value is estimated.

[0020] Furthermore, step S3 includes:

[0021] First, the signal shape parameters are estimated based on the received mixed signal and interference signal, and the final shape parameter estimation result is obtained. Then, the weight function of the received mixed signal is calculated based on the semi-blind source separation algorithm to obtain the optimal estimate of the target communication signal.

[0022] Furthermore, the signal shape parameter estimation algorithm is as follows: First, the received mixed signal and interference signal are transformed by STFT to obtain their respective time-frequency matrices. Then, the time-frequency matrix of the mixed signal is processed, and the shape parameters are estimated by deep learning neural network method and numerical calculation method respectively, and the final shape parameter estimation result is obtained.

[0023] Furthermore, the semi-blind source separation of the signal is achieved by: calculating the weight function of the received mixed signal, and iteratively updating the covariance matrix and separation matrix of the signal accordingly, finally calculating the time-frequency representation matrix of the target estimated signal, and obtaining the time-domain form of the estimated target communication signal through iSTFT.

[0024] Furthermore, the signal follows a generalized Gaussian distribution in the time domain.

[0025] The beneficial effects of this invention are as follows:

[0026] This invention improves the accuracy of the algorithm by employing techniques such as the structural design of the neural network in the signal shape parameter estimation algorithm, the joint estimation method of the shape parameter of the generalized Gaussian distribution signal based on numerical analysis and neural network, the semi-blind source separation algorithm based on shape parameter estimation and AuxIVA, and the communication signal anti-interference enhancement algorithm based on shape parameter estimation and AuxIVA. This effectively enhances the anti-interference capability of the signal against co-band interference signals that cause interference in wireless communication. Attached Figure Description

[0027] Figure 1 This is a schematic diagram illustrating an application scenario of the communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to the present invention.

[0028] Figure 2 This is a schematic diagram of a neural network architecture used for signal shape parameter estimation;

[0029] Figure 3 This is a schematic flowchart of the communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to the present invention;

[0030] Figure 4 This is a comparison chart of the frequency band occupancy of two types of communication signals;

[0031] Figure 5 This is a comparison chart of BER before and after signal enhancement under different SIRs. Detailed Implementation

[0032] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0034] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0035] The following combination Figures 1-5 Specific embodiments of the present invention will be described in detail below. It should be understood that the specific embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit the present invention.

[0036] Traditional blind source separation algorithms are often used in scenarios involving multiple unknown signal sources. The goal is to separate each signal source from the mixed signal, aiming for high separation accuracy across all sources. However, in practical communication systems, especially point-to-point communication, the receiver often only focuses on the purity of a specific communication signal, without considering the separation accuracy of other sources. Furthermore, the receiver is usually not entirely unaware of interfering signals; it often possesses some prior information, such as data collected when friendly communication signals are silent, or information about the frequency band, modulation scheme, and waveform of the interfering signal obtained through signal parameter estimation techniques. The only unknown is the channel through which the interfering signal propagates. Therefore, in practical communication systems, a semi-blind source separation algorithm, evolved from the traditional blind source separation algorithm, can be used to estimate the channel function of the interfering signal from the mixed signal. This effectively reduces or even eliminates the influence of the interfering signal in the mixed signal, significantly enhancing the anti-interference capability of the target communication signal.

[0037] The application scenarios targeted by the method of the present invention are as follows: Figure 1 As shown, this invention primarily considers interference signal cancellation algorithms. For example, electromagnetic fields, currently widely affected by resource pollution and spectrum interference, are extensively used in radio wave communication. Due to the intensive use of radio frequency signal resources, mutual interference between communication signals in adjacent frequency bands has become commonplace; even worse, in certain specific scenarios, unknown devices may maliciously transmit signals in the same frequency band to cause interference. The communication scenario targeted by this algorithm is a typical point-to-point communication, where both the transmitting and receiving devices are radio frequency signal processing devices with only one antenna each. The time-domain transmitted signal from the transmitting end is... The time-domain received signal obtained by the receiving end is There are interfering signal sources in the area, and the emitted time-domain interference signal is... The impulse response of the interference signal after passing through the multipath channel is The transmitted signal and the interference signal are additively mixed at the receiving end, and the received signal at the receiving end can be represented as... ,in This is a temporal convolution. The ultimate goal of this algorithm is to perform convolution on communication signals. and interference signal multipath channel In unknown situations, by receiving signals and prior information about the interference signal For communication signals Enhancement and recovery are carried out to improve the overall anti-interference capability of the communication system.

[0038] Based on this, the present invention proposes a communication signal anti-interference method based on shape parameter estimation and semi-blind source separation, the method comprising the following steps:

[0039] S1. The signal receiving device acquires the received signal through an antenna. The received signal For including transmitted signals from the signal transmitting device Interference signals from other sources .

[0040] S2. Combining deep learning and numerical analysis methods, the estimated values ​​of the signal shape parameters of the received signal are obtained; the signal shape parameters are the shape parameters in the probability density function of the generalized Gaussian distribution to which the signal follows, i.e., the parameters. .

[0041] S3. Receive signal and interference signals The signal shape parameter estimate is input into the signal processing system, which uses a semi-blind source separation algorithm based on shape parameter estimation and AuxIVA to obtain the target communication signal.

[0042] This invention first estimates the shape parameters of the signal, and then applies a semi-blind source separation algorithm based on the estimated shape parameter values ​​to obtain the estimated target communication signal. After obtaining the target communication signal, the receiver performs its original downsampling, mixing, demodulation, and decoding operations. The target communication signal finally obtained in step S3 is the transmitted signal after system processing. The approximate estimated signal.

[0043] The anti-interference method proposed in this invention is a semi-blind source separation algorithm combining AuxIVA and deep learning. AuxIVA-type algorithms, based on ICA algorithms, extend the signal processing domain from the time domain to the time-frequency domain, effectively improving the correlation between frequency components within the same signal to solve the permutation ambiguity problem during signal separation. An auxiliary function is introduced to improve the convergence speed. This algorithm requires the use of shape parameters. Shape parameters The value of is uncertain. Existing algorithms often use empirical values, such as 0.4, when employing this parameter, which limits the algorithm's accuracy. Therefore, in step S2 of this invention, the shape parameter... To accurately determine the value of shape parameters, a shape parameter estimation algorithm combining deep learning and numerical analysis is proposed. This improvement in parameter estimation accuracy enhances the algorithm's precision and computational accuracy, thereby increasing the accuracy of target communication signal extraction and strengthening anti-interference capabilities. The signal shape parameter estimation is implemented as follows: First, the received mixed signal and interference signal are subjected to STFT transformation to obtain their respective time-frequency matrices. Then, the time-frequency matrix of the mixed signal is processed, and shape parameter estimation is performed using both deep learning neural network methods and numerical calculation methods to obtain the final shape parameter estimation result. Specifically, the following steps are included:

[0044] S2.1 uses deep learning to obtain the estimated values ​​of the first signal shape parameters. Specifically, the neural network architecture used for signal shape parameter estimation is as follows: Figure 2 As shown, signal shape parameter estimation is based on the signal's time-frequency matrix. The overall network architecture includes two convolutional neural network (CNN) layers, two average pooling layers, a flattening layer, and a fully connected layer. The kernel size of both the CNN layers and the pooling layers is 3×3. Before the algorithm runs, a set of random signals with specific shape parameter values ​​is generated for network training. After training, the network is then used for signal shape parameter estimation.

[0045] S2.2 uses numerical analysis to obtain the estimated values ​​of the second signal shape parameters. Also based on the time-frequency representation matrix of the signal, the algorithm flow is shown below, including the following steps:

[0046] S2.2.1 The complex signal kurtosis (CSK) value of the signal is calculated using the time-frequency matrix as follows:

[0047] ;

[0048] in, Represents time-domain signal The time-frequency representation, This indicates that the mean of the signal is calculated.

[0049] S2.2.2 Based on CSK value and shape parameters The relationship of values, to The value is estimated. The specific relationship is shown below:

[0050] ;

[0051] At this point, the shape parameter estimation using numerical analysis methods has been completed.

[0052] S2.3 The average of the first shape parameter estimate and the second shape parameter estimate is taken as the final result of the signal parameter estimate.

[0053] In step S3, the communication signal anti-interference enhancement algorithm proposed in this invention includes a shape parameter estimation algorithm and a semi-blind source separation algorithm based on AuxIVA (Auxiliary Function-Based Independent Vector Analysis). The shape parameter estimation algorithm includes deep learning and numerical computation methods, and the final result of the shape parameter estimation is the average of the deep learning and numerical computation results. The flowchart of the overall algorithm is as follows: Figure 3 As shown. In Figure 3 In this process, the received, interfered communication signal undergoes two steps: signal shape parameter estimation and semi-blind source separation, after which the optimal estimate of the target communication signal can be obtained. The goal of the semi-blind source separation algorithm is to achieve semi-blind source separation. This is achieved by calculating the weight function of the received mixed signal and iteratively updating the signal's covariance matrix and separation matrix accordingly. Finally, the time-frequency representation matrix of the estimated target signal is calculated, and the time-domain form of the estimated target communication signal is obtained through iSTFT.

[0054] Specifically, step S3 includes:

[0055] Signal shape parameter estimation, as detailed below:

[0056] The time-domain received signal is represented by a time-frequency matrix. Time-domain interference signals and the estimated time-domain target communication signal By performing a Short-Time Fourier Transform (STFT), the time-frequency representation matrices of the three signals are obtained:

[0057] ;

[0058] Accordingly, the generation expression of the mixed signal in the time-frequency domain can be expressed as:

[0059] ;

[0060] in, The first 1000 STFT results of the interference signal The influence coefficient of the frame signal on the current frame signal; The index is the coefficient subscript number, specifically the first index in the STFT result of the interference signal. The index number of the coefficient that affects the STFT result of the received signal in the current frame; The influence coefficient of the current frame in the STFT result of the interference signal on the current frame in the STFT result of the received signal; This is the original transmitted signal; This refers to the number of adjacent frames that need to be considered simultaneously. By constructing an extended matrix, representations of the generalized mixed signal and the generalized target signal are obtained: (The purpose of this step is to extend the computation process of the semi-blind source separation algorithm, transforming it into a form that can be directly applied to the blind source separation algorithm.)

[0061] ;

[0062] In the STFT results of the interference signal, the current frame and the most recent frame The vector consisting of the STFT result of the frame pair received signal and the influence coefficient value on the current frame.

[0063] It is known that the extended matrix and All The column vector. Define the generalized target signal. The calculation method is as follows: (The meaning of this formula is: for the extended vectors x and y, the separation matrix W after iterative estimation is applied to calculate the extended vector x.)

[0064] ;

[0065] in, It is the time-frequency representation of the separation matrix, which is a... The complex field matrix can be decomposed into:

[0066] ;

[0067] in It is a length of A column vector consisting entirely of zeros. It is A 3D identity matrix, the goal is to... The first line Estimation is performed. The parameters used in this algorithm are the received signals in the time domain. Interference signals in the time domain The main calculation result obtained is the target estimation signal in the time domain. The purpose of this step is to estimate the final target signal, therefore this algorithm is a global algorithm.

[0068] A semi-blind source separation algorithm based on the AuxIVA algorithm. The algorithm uses the received signal in the time-frequency domain as the parameter. Also used is a pre-trained algorithm for estimating shape parameters. The neural network primarily obtains shape parameters by combining neural network and numerical calculation methods. The purpose of this step is to estimate the shape parameters by applying the shape parameter estimation algorithm proposed in this invention. The details are as follows:

[0069] In the AuxIVA algorithm, the signal follows a generalized Gaussian distribution (GGD) in the time domain, and its probability density function is expressed as follows:

[0070] ;

[0071] in, The symbol representing the independent variable in the probability density expression; signal shape parameter. This is the key to signal parameter estimation in this algorithm; the signal shape parameters are obtained through step S2. Estimated value; and The mean and variance of the signal are parameters. , The calculation method is as follows:

[0072] ;

[0073] It is the gamma function, a general function used to represent the probability density function of GGD, and its expression is:

[0074] ;

[0075] For an integral variable, It is the natural logarithm. It is the negative power of the natural logarithm.

[0076] Suppose the signal obeys the complex field spherical GGD, that is:

[0077] ;

[0078] It is a temporary representation of the target signal and can refer to any time-domain signal of the estimated target signal.

[0079] Define the comparison function and the weight function as follows:

[0080] ;

[0081] in, For signal The 2-norm; These are meaningless identifiers. yes The first derivative; ; , Represents the L2 norm. These are auxiliary variables in AuxIVA. The comparison function and weighting function are used to iteratively calculate the covariance matrix. And calculate the separation matrix. .

[0082] The weight covariance matrix is ​​calculated as follows: and separation matrix :

[0083] ;

[0084] in, For the weight function, This is the conjugate transpose of the matrix. The vector is Multiply by its conjugate transpose and take the mean in the time domain; It is the forgetting factor. Through iteration on the separation matrix... Perform the update, and finally retrieve the matrix. The first row is used as the final separation matrix. The first line This completes the separation matrix. The estimate.

[0085] Through the estimated separation matrix Calculate the time-frequency matrix representation of the source signal The time-domain representation of the source signal is obtained through inverse short-time Fourier transform (ISTFT). The source signal here refers to the optimal estimate of the transmitted signal, i.e., the target communication signal.

[0086] Thus, the semi-blind source separation algorithm based on the AuxIVA algorithm is completed. This algorithm can be used in point-to-point communication systems to enhance the anti-interference capabilities of communication signals.

[0087] The algorithm of this invention was tested in a mixed signal of two wireless communication signals. Binary Phase Shift Keying (BPSK) and Gaussian Filtered Minimum Shift Keying (GMSK) are two widely used communication modulation methods, and therefore were used in the technical effect test. The sampling frequency of both communication signals was 40MHz, the radio frequency was 10MHz, the intermediate frequency was 1MHz, and the length of the modulated signal for both was approximately 1 second. The code rate of the BPSK modulated signal was 4.9kbps, and the code rate of the GMSK modulated signal was 2.4kbps. The frequency band occupied by the two communication radio frequency signals is compared as follows: Figure 4 As shown, both types of signals occupy the frequency band within the range of 10.987MHz to 11.013MHz, indicating a high degree of overlap in their communication frequency band occupancy. When both communication signals are transmitted simultaneously, the BPSK communication signal at a certain power will severely interfere with the demodulation of the GMSK signal. Under different signal-to-interference ratios (SIR), the bit error rate (BER) of the interfered GMSK signal before and after applying the signal enhancement algorithm proposed in this invention is compared to... Figure 5 As shown, when the SIR is less than 10dB, the interference signal significantly affects the demodulation of the communication signal, resulting in a high BER. However, after processing by the signal enhancement algorithm proposed in this invention, the BER is effectively improved and can be maintained at a very low level. Conversely, when the SIR is high and reaches above 10dB, the interference signal is weak, and the effect of applying the algorithm of this invention is not significant. These results demonstrate that the method proposed in this invention has effective anti-interference capability against co-frequency band interference signals that cause interference in wireless communication.

[0088] Any process or method described in the flowcharts of this invention or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, which can be implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device. The computer-readable medium can be any medium containing a program for storage, communication, propagation, or transmission for use by the execution system, apparatus, or device, including read-only memory, magnetic disks, or optical disks.

[0089] In the description of this specification, references to terms such as "embodiment," "example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, those skilled in the art can combine or combine the different embodiments or examples described in this specification and the features therein without causing contradiction.

[0090] While embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and alterations to the above embodiments within the scope of the present invention.

Claims

1. A communication signal anti-interference method based on shape parameter estimation and semi-blind source separation, characterized in that, The method includes the following steps: S1. The signal receiving device acquires a received signal through an antenna, the received signal being a mixed signal including a transmitted signal from the signal transmitting device and interference signals from other sources; S2. Combine deep learning and numerical analysis methods to obtain signal shape parameters. Estimated value; S3. Combine the received signal, interference signal, and signal shape parameters. The estimated value is input into the signal processing system, which uses a signal shape parameter estimation algorithm and an AuxIVA-based semi-blind source separation algorithm to obtain the target communication signal. Obtain signal shape parameters The steps for estimating the value are as follows: The time-domain received signal is represented by a time-frequency matrix. Time-domain interference signals and the estimated time-domain target communication signal By performing a short-time Fourier transform (STFT), the time-frequency matrices of the three signals are obtained: ； ； ； Accordingly, the generation expression of the mixed signal in the time-frequency domain can be expressed as: ； in, The first 1000 STFT results of the interference signal The influence coefficient of the frame signal on the current frame signal; For coefficient subscript numbers, This is the original transmitted signal; It is the number of adjacent frames that need to be considered simultaneously; In the STFT results of the interference signal, the current frame and the most recent frame STFT results of the frame-to-received signal: ； A semi-blind source separation algorithm based on the AuxIVA algorithm uses the received signal in the time-frequency domain as the parameter. Simultaneously, using pre-trained parameters for estimating shape parameters The neural network obtains shape parameters that are obtained simultaneously through neural network and numerical calculation methods. The estimated values ​​are obtained by using a shape parameter estimation algorithm. The estimated values ​​are as follows: In the AuxIVA algorithm, the signal follows a generalized Gaussian distribution in the time domain, and its probability density function is expressed as follows: ； in, This is the symbol representing the independent variable in the probability density expression. and The mean and variance of the signal are parameters. , It is a gamma function; Calculate the weight covariance matrix and separation matrix : ； ； in, For the weight function, This is the conjugate transpose of the matrix. To make vector Multiply by its conjugate transpose and take the mean in the time domain; It is the forgetting factor; through iteration on the separation matrix Perform the update, and finally retrieve the matrix. The first row is used as the final separation matrix. The first line This completes the separation matrix. The estimate; It is the time-frequency representation of the separation matrix, which is a The complex field matrix is ​​decomposed into: ； in It is a length of A column vector consisting entirely of zeros. It is A 3D identity matrix, the goal is to... The first line Make an estimate; Through the estimated separation matrix Calculate the time-frequency matrix representation of the source signal The time-domain representation of the source signal is obtained through inverse short-time Fourier transform. The source signal is the optimal estimate of the transmitted signal, i.e., the target communication signal.

2. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 1, characterized in that, Step S2 specifically includes the following steps: S2.1 uses deep learning to obtain the estimated values ​​of the shape parameters of the first signal; S2.2 The estimated values ​​of the shape parameters of the second signal are obtained by numerical analysis; S2.3 The average of the first signal shape parameter estimate and the second signal shape parameter estimate is taken as the final result of the signal parameter estimate.

3. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 2, characterized in that, The specific implementation method of step S2.1 is as follows: establish a neural network for signal shape parameter estimation. The signal shape parameter estimation is based on the time-frequency matrix of the signal. The overall network architecture includes 2 convolutional neural network layers, 2 average pooling layers, a flattening layer, and a fully connected layer.

4. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 3, characterized in that, The kernel size of both the CNN layer and the pooling layer is 3×3.

5. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 4, characterized in that, Before running the neural network algorithm for signal shape parameter estimation, a set of random signals with specific shape parameter values ​​is first generated for network training. After training, the network is then used for signal shape parameter estimation.

6. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 2, characterized in that, The specific implementation of step S2.2 is as follows: Also based on the time-frequency representation matrix of the signal, the algorithm flow includes the following steps: S2.2.1 First, the kurtosis value of the complex signal is calculated using the time-frequency matrix; S2.2.2 Then, based on the CSK value and signal shape parameters... The relationship between values, for The value is estimated.

7. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 1, characterized in that, Step S3 includes: First, the signal shape parameters are estimated based on the received mixed signal and interference signal, and the final shape parameter estimation result is obtained. Then, the weight function of the received mixed signal is calculated based on the semi-blind source separation algorithm to obtain the optimal estimate of the target communication signal.

8. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 7, characterized in that, The signal shape parameter estimation algorithm is as follows: First, the received mixed signal and interference signal are transformed by STFT to obtain their respective time-frequency matrices. Then, the time-frequency matrix of the mixed signal is processed, and the shape parameters are estimated by deep learning neural network method and numerical calculation method respectively, and the final shape parameter estimation result is obtained.

9. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to claim 7, characterized in that, The semi-blind source separation of the signal is achieved by: calculating the weight function of the received mixed signal, and iteratively updating the covariance matrix and separation matrix of the signal accordingly, finally calculating the time-frequency representation matrix of the target estimated signal, and obtaining the time-domain form of the estimated target communication signal through iSTFT.

10. The communication signal anti-interference method based on shape parameter estimation and semi-blind source separation according to any one of claims 1-9, characterized in that, The signal follows a generalized Gaussian distribution in the time domain.

Images