Wideband sensing method based on MIMO MWC sub-Nyquist sampling structure

Abstract

The application relates to a wide-band sensing method based on a MIMO-based MWC sub-Nyquist sampling structure, which comprises the following steps: a MIMO-based MWC wide-band sampling front end is used to model a signal in an mCSL matrix; a static covariance matrix is constructed by using the correlation between column vectors of the matrix; subspace decomposition is performed on the static covariance matrix to obtain eigenvalues corresponding to signals and eigenvalues corresponding to noises, and noise reconstruction is performed; according to the orthogonality of the signals and the noises, a calculation vector based on the mCSL matrix structure is obtained based on the reconstructed noises, and the calculation vector is sorted; support sets corresponding to the first K values of the calculation vector are used as negative frequency points where PU users are located, and all center frequency points of the PU users are obtained according to the symmetry of real signal spectrum. The application can improve the spectrum recovery effect.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a wideband sensing method based on a MIMO-based MWC under-Nyquist sampling structure. Background Technology

[0002] With the popularization and development of communication technology, the massive increase in demand for wireless devices has led to increasingly scarce spectrum resources, with less and less unlicensed spectrum available. To utilize spectrum resources more rationally and avoid spectrum waste, cognitive radio technology has become a primary solution. In broadband sensing, due to the wide range of the sensing spectrum, the Nyquist rate sampling method places high demands on the sampling rate of the ADC, which traditional ADCs struggle to meet. Furthermore, the large amount of data generated after broadband sampling puts enormous pressure on the hardware's processing and computation, causing the entire cognitive radio (CR) system to consume a significant amount of energy. To achieve under-Nyquist sampling, a method such as... Figure 1 The diagram shows a Modulated Wideband Converter (MWC) structure for multiple-input multiple-output (MIMO).

[0003] Existing patent document CN113596850A discloses a broadband spectrum sensing method suitable for MWC sub-Nyquist sampling structures. This method shifts the spectrum of the received broadband signal to near the baseband. The shifted signal not only folds the original signal information but also utilizes a continuous-to-finite (CTF) module for spectrum recovery. This structure significantly reduces the sampling bandwidth and also greatly reduces the sampling pressure on the ADC. However, this patent document only addresses a single MWC structure and does not consider the performance improvements brought about by the correlation between channels in multi-antenna MIMO MWC structures.

[0004] Subspace learning has been widely applied in signal processing. Subspace learning methods combined with MIMO antennas have seen significant development in spectrum sensing. These methods primarily construct overdetermined equations by building a statistical covariance matrix, and then solve these equations to achieve spectrum sensing. Subspace learning methods have evolved alongside narrowband spectrum sensing to broadband sensing. Currently, mainstream compressed subspace learning methods are mainly divided into orthogonal matching pursuit (OMP) algorithms, multiple signal classification (MUSIC) algorithms, and convex optimization methods. While convex optimization methods are highly effective, they suffer from excessive computational complexity and high computational resource consumption. The mCSL algorithm proposed by Tierui Gong et al. not only integrates the utilization of spatial correlation in MIMO but also introduces a compressed subspace method, reducing the overhead of the CTF module and improving sensing performance. However, because its mCSL structure is based on the Orthogonal Matching Pursuit (OMP) algorithm, while it performs well in sensing low-noise signals, it may suffer from a high number of iterations and computational complexity. Compared to the other two algorithms, the MUSIC algorithm has lower computational complexity and is easier to implement in hardware. Furthermore, MIMO utilizes spatial correlation to compensate for its lack of noise robustness. However, it also shares the drawbacks of traditional MUSIC algorithms, which are mostly derived in noise-free environments and perform poorly in noisy conditions. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a wideband sensing method based on MIMO-based MWC under-Nyquist sampling structure, which can improve the spectrum recovery effect.

[0006] The technical solution adopted by this invention to solve its technical problem is: to provide a broadband sensing method based on MIMO-based MWC under-Nyquist sampling structure, comprising the following steps:

[0007] The signal is modeled using an mCSL matrix by a MIMO-based MWC wideband sampling front-end.

[0008] Construct a static covariance matrix using the correlation between the column vectors of the matrix;

[0009] The static covariance matrix is ​​subspace decomposed to obtain the eigenvalues ​​corresponding to the signal and the eigenvalues ​​corresponding to the noise, and then the noise is reconstructed.

[0010] Based on the orthogonality of signal and noise, a computational vector based on the mCSL matrix structure is obtained from the reconstructed noise, and the computational vector is sorted.

[0011] The set of support points corresponding to the first K values ​​of the calculated vector is taken as the negative frequency points of the PU user. Based on the symmetry property of the real signal spectrum, all center frequency points of the PU user are obtained.

[0012] The mCSL matrix modeling representation is as follows: ,in, For mCSL matrix, The MIMO channel description matrix is ​​represented as follows: , This is the correlation coefficient matrix at the receiving end. The correlation coefficient matrix at the transmitter is... Represents the correlation matrix of random antennas. For the received signal, Represents the perception matrix, This is the noise signal for AWGN.

[0013] The static covariance matrix is ​​expressed as follows: ,in, It is a static covariance matrix. Represents the perception matrix, Indicates the number of PU users. For the gain of the antenna and receiver, The correlation coefficient at the receiving end. It represents the mathematical expectation.

[0014] The subspace decomposition of the static covariance matrix to obtain eigenvalues ​​corresponding to the signal and eigenvalues ​​corresponding to the noise, followed by noise reconstruction, includes:

[0015] The eigenspace decomposition of the static covariance matrix is ​​calculated as follows: ,in, For feature vectors, The characteristic value corresponding to the signal. These are the eigenvalues ​​corresponding to the noise. Represents the block singular value decomposition function;

[0016] The signal's corresponding eigenvalues ​​and the noise's corresponding eigenvalues ​​are classified using the number of PU users and noise levels. Represented as: eigenvalues ​​corresponding to noise Represented as: , This refers to the number of antennas.

[0017] Through feature vectors The noise is reconstructed, and the reconstructed noise is... Represented as: , This represents the final singular value obtained from the calculation.

[0018] The computation vector pass The calculation yielded that, For averaging, For the noise of reconstruction, This represents the perception matrix.

[0019] Beneficial effects

[0020] By adopting the above-mentioned technical solutions, this invention has the following advantages and positive effects compared with existing technologies: To better utilize spatial diversity, this invention considers MIMO as the hardware foundation and proposes an improved music algorithm based on the mCSL structure, leveraging the spatial correlation advantages brought by the mCSL structure. Simultaneously, to further reduce the algorithm's weight, the computation vector of the improved music algorithm is provided. The proposed algorithm outperforms the mainstream SOMP algorithm, effectively resisting noise, increasing detection probability, reducing computational complexity, and improving the performance of broadband sensing. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the MWC under-Nyquist sampling front-end based on MIMO;

[0022] Figure 2 This is a flowchart of the broadband sensing method according to an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the mCSL sensing matrix structure under the MIMO structure in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram illustrating the recovery effect of an embodiment of the present invention;

[0025] Figure 5 This is a graph showing the relationship between the number of antennas with an SNR of 5dB and the detection probability in an embodiment of the present invention.

[0026] Figure 6 This is a graph showing the relationship between the number of antennas with an SNR of 8dB and the detection probability in an embodiment of the present invention.

[0027] Figure 7 This is a graph showing the relationship between the number of antennas and the detection probability in an embodiment of the present invention with an SNR of 10dB.

[0028] Figure 8 This is a graph showing the relationship between the number of antennas and the detection probability in an embodiment of the present invention with an SNR of 15dB. Detailed Implementation

[0029] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0030] This invention relates to a broadband sensing method based on a MIMO-based MWC under-Nyquist sampling structure. The method proposes a spectrum sensing algorithm based on an improved MUSIC matrix compressive subspace learning (mCSL) structure. This method utilizes the spatial correlation between the autocorrelation coefficients of MIMO signals to decompose the compressive subspace learning algorithm, providing the computational vector of the algorithm. It achieves signal reconstruction without optimization or iteration, reducing receiver complexity, and the computational complexity does not increase with the maximum number of unknown signals. Figure 2 As shown, the specific steps include:

[0031] (1) Signal modeling is performed using a MIMO MWC wideband sampling front-end, i.e., the signal is modeled using an mCSL matrix using a MIMO MWC wideband sampling front-end. Consider a scenario where K users exist simultaneously in a certain area. Consider the existence of multi-band signals within the Hilbert space. It is the superposition of the narrowband signals of K PU users in the transmitted wideband, which can be expressed as: .

[0032] The signal received by each antenna at the receiving end contains not only the sum of signals from K users, but also additive white Gaussian noise with a mean of 0 and a variance of . Therefore, the signal received at one end is represented as: .

[0033] in, This represents the total signal received by M antennas at time t within one sampling period. This represents the channel gain matrix between M antennas and K PUs, i.e., the MIMO channel description matrix, which greatly improves the correlation between multiple antennas. This represents the sum of K user signals. This represents the noise signal of the AWGN. The noise experienced by the K user signals received by each antenna follows the same distribution but is independent of each other. The MIMO channel model adopts the Kronecker model, and thus the expression for its channel description matrix H is obtained, which can be expressed as: .

[0034] in, This is the correlation coefficient matrix at the receiving end. Let be the correlation coefficient matrix of the transmitter. The k-th diagonal element of the transmitter matrix conforms to... The correlation coefficient matrix at the receiving end It consists of the correlation coefficients between different antennas. The correlation matrix of random antennas can be represented as: The mean is 0 and the variance is . In a multi-antenna MIMO MWC structure system, each antenna carries an identical MWC structure. To reduce computational complexity, the channels of the M MIMO MWCs are processed sequentially. Therefore, the k-th sensing matrix represents the sensing matrix constructed from the k-th channel of all M MWCs. Each i-th row of the sensing matrix represents the undersampled information of the MWC channel corresponding to the i-th antenna, and its data structure is as follows... Figure 3 As shown. Therefore, the mCSL matrix can be represented as: Where the superscript T denotes the transpose matrix, This represents the perception matrix in the MWC sampling structure.

[0035] (2) Construct a static covariance matrix using the correlation between the column vectors of the matrix. , is represented as: Where the superscript H denotes the Hermilte matrix, Here, represents the gain of the antenna and receiver; the superscript * indicates the adjoint matrix. The correlation coefficient at the receiving end. It represents the mathematical expectation.

[0036] (3) Calculate the static covariance matrix The characteristic subspace decomposition of can be expressed as: ,in, For feature vectors, The characteristic value corresponding to the signal. These are the eigenvalues ​​corresponding to the noise. This represents the block singular value decomposition function.

[0037] (4) Classify the characteristic values ​​corresponding to the signal and the characteristic values ​​corresponding to the noise using the number of PU users and noise. Represented as: eigenvalues ​​corresponding to noise Represented as: By reconstructing the signal using eigenvalues ​​and eigenvectors, we can obtain expressions for both the signal and noise. , .

[0038] (5) Based on the orthogonality of signal and noise, a computational vector based on the mCSL matrix structure is given, that is, the signal support set is obtained for the subspace. Combined with the CTF module, since the signal subspace and the noise subspace are orthogonal, when there is noise in the signal, the noise eigenvalues ​​form a noise spanning space, making the computational vector... Among them, the vector is calculated. The expression is: .in, This is for averaging. Due to orthogonality, it offers the advantages of non-iterative and fast computation, improving the accuracy and complexity of traditional music calculations. Specifically, it calculates vectors... The value is not 0, therefore the calculation vector Sort them.

[0039] (6) The support set corresponding to the first K values ​​of the calculated vector is taken as the negative frequency point of the PU user. Based on the symmetry property of the real signal spectrum, all center frequencies of the PU user are obtained. Specifically, the unknown broadband signal is estimated. negative frequency part support set Calculate the positive frequency support set of the unknown broadband signal as follows: ,in, This represents the length of the total sampled values, and the estimated complete support set is... Therefore, this implementation method only calculates (L+1) / 2 vectors Z, and directly calculates the other N support sets, thus reducing computational complexity. Finally, the spectrum sensing results are output.

[0040] This implementation presents the performance metrics of the mCSL structure under different parameters. The recovery performance of the improved music algorithm for the mCSL structure is shown, and the changes in detection probability under different numbers of antennas and signal-to-noise ratios are compared.

[0041] (1) Noise performance of the improved music algorithm under the mCSL structure: The fading channel through which the signal passes during propagation is considered to be a Gaussian channel. The power spectral density of the original signal is plotted. The horizontal axis of the power spectrum represents time, and the vertical axis represents the analog frequency, i.e., 5GHz. The color represents the power spectral density corresponding to the frequency. The suppression of noise and spectrum perception under the time-frequency image are shown. In order to reflect the superiority of the music algorithm and reduce hardware complexity, the number of MIMO antennas M=3. Figure 4The image demonstrates the noise immunity performance of the improved music algorithm with a detection probability of 100% in the mCSL structure at an SNR of -15dB. It can be seen that in the absence of noise, the first time-frequency images (TFI) show three distinct bright lines corresponding to different sub-frequency bands of the three PU user signals, with the line width equal to the bandwidth of the PU user signal sub-band. When noise is present in the propagation environment, the PU user signal is observed to be submerged in noise in the second TFI image. In contrast, the PU user signal is not observable in the noise-free case. The third TFI image shows that the improved music algorithm in the mCSL structure effectively filters noise, exhibiting clear energy characteristics, and the power resolution remains essentially the same.

[0042] (2) Compared with the mainstream SOMP recovery algorithm, the algorithm using the improved music significantly outperforms the mainstream SOMP algorithm in terms of detection probability after SNR>=5dB. The horizontal axis represents the number of antennas, and the vertical axis represents the detection probability. Different structural algorithms are shown below. Figure 5-8 As shown, the algorithm performance improves with the increase in the number of antennas, indicating that the increased number of antennas makes the overdetermined equations constructed by the improved music algorithm more stable and further enhances the utilization of antennas. When the noise is high, increasing the number of antennas significantly improves the detection performance. When the noise is low, the contribution of the number of antennas to the detection probability gradually decreases after a certain point. The proposed improved music algorithm not only maintains the high recovery rate of the music algorithm when the signal noise is low, but also has lower computational complexity. The SOMP algorithm compared to it has been proven to be superior to other traditional OMP algorithms in the work of Tierui Gong et al., and the comparative experiments also show that the two structurally improved music algorithms are significantly better than the SOMP algorithm.

[0043] It is easy to see that, in order to better utilize spatial diversity, this invention considers MIMO as the hardware foundation and proposes an improved music algorithm based on the mCSL structure, leveraging the spatial correlation advantages brought by the mCSL structure. Furthermore, to further reduce the algorithm's weight, the computation vector of the improved music algorithm is provided. The proposed algorithm outperforms the mainstream SOMP algorithm, effectively resisting noise, increasing detection probability, reducing computational complexity, and improving the performance of broadband sensing.

Claims

1. A wideband perception method based on MIMO-based MWC sub-Nyquist sampling structure, characterized in that, Includes the following steps: The signal is modeled using an mCSL matrix by a MIMO-based MWC wideband sampling front-end. Construct a static covariance matrix using the correlation between the column vectors of the matrix; Subspace decomposition of the static covariance matrix yields eigenvalues ​​corresponding to the signal and eigenvalues ​​corresponding to the noise, followed by noise reconstruction, including: a eigen-subspace decomposition of the static covariance matrix is computed, denoted as: where, are the eigenvectors, are the eigenvalues corresponding to the signal, are the eigenvalues corresponding to the noise, denotes a block singular value decomposition function; The number of PU users is used to classify the eigenvalue corresponding to the signal and the eigenvalue corresponding to the noise, and the eigenvalue corresponding to the signal is expressed as: , the eigenvalue corresponding to the noise is expressed as: , is the number of antennas; by the eigenvectors reconstructing the noise, the reconstructed noise is represented as: , denotes the last singular value obtained by the calculation; wherein, denotes the number of PU users; Based on the orthogonality of signal and noise, a computational vector based on the mCSL matrix structure is obtained from the reconstructed noise, and the computational vector is sorted. The set of support points corresponding to the first K values ​​of the calculated vector is taken as the negative frequency points of the PU user. Based on the symmetry property of the real signal spectrum, all center frequency points of the PU user are obtained.

2. The method of claim 1, wherein the MIMO-based MWC sub-Nyquist sampling structure wideband sensing method is characterized by, The mCSL matrix modeling representation is as follows: ,in, For mCSL matrix, The MIMO channel description matrix is ​​represented as follows: , This is the correlation coefficient matrix at the receiving end. The correlation coefficient matrix at the transmitter is... Represents the correlation matrix of random antennas. For the received signal, Represents the perception matrix, This is the noise signal for AWGN.

3. The broadband sensing method based on MIMO-based MWC under-Nyquist sampling structure according to claim 1, characterized in that, The static covariance matrix is represented as: wherein, is a static covariance matrix, denotes a perception matrix, is a gain of an antenna and a receiving end, is a correlation coefficient of a receiving end, denotes a mathematical expectation.

4. The broadband sensing method based on MIMO-based MWC under-Nyquist sampling structure according to claim 1, characterized in that, The calculation vector By is calculated, wherein is an average operation, is a reconstructed noise, denotes a perception matrix.

Images