A multipath time delay estimation method based on improved MUSIC algorithm

Abstract

This invention discloses a multipath delay estimation method based on an improved MUSIC algorithm. The implementation steps are as follows: calculate the normalized cross-spectral sequence h[k] of the received signal under multipath conditions; and use the spatial smoothing idea to divide h[k] into multiple overlapping subsequences h[k]. q Take the conjugate data x of each subsequence q Using h q and x q The improved covariance matrix is ​​calculated and then subjected to eigenvalue decomposition to obtain noise eigenvalues ​​and their corresponding noise eigenvectors. The noise eigenvalues ​​are then corrected, and the corrected eigenvalues ​​are used to improve the MUSIC time delay estimation spectrum. Finally, spectral peak search is performed on the estimated spectrum to obtain the multipath time delay value. This invention effectively utilizes the conjugate information and noise eigenvalue information of the measurement data, enabling high-precision estimation of multipath time delay under conditions of small sample size and low signal-to-noise ratio.

Description

TECHNICAL FIELD

[0001] The present application relates to signal processing, in particular a multipath time delay estimation method based on improved MUSIC algorithm. BACKGROUND

[0002] The positioning technology based on time difference of arrival (TDOA) estimates the position of the target by measuring the time difference of signal arrival at each receiver, which has the advantages of simple architecture, high positioning accuracy and strong flexibility, and has important significance for the application in modern electronic warfare. The accuracy of time delay estimation directly affects the accuracy of target positioning, so accurate estimation and resolution of the time delay of the received signal is the key part of TDOA positioning technology. The resolution capability of the traditional time delay estimation method based on correlation analysis is limited by the bandwidth, and the performance deteriorates sharply in a multipath environment. Therefore, the super-resolution multipath time delay estimation algorithm which breaks through the correlation time-Rayleigh limit is the current research focus.

[0003] The super-resolution multipath time delay estimation algorithm is mainly divided into three categories: maximum likelihood estimation based algorithm, subspace algorithm and sparse optimization algorithm. Among them, the multiple signal classification (MUSIC) algorithm of subspace class was proposed by Schmid in 1986, which was initially used to realize the super-resolution estimation of direction of arrival (DOA) (R O Schmidt. Multiple emitter location and signal parameter estimation[J]. IEEE Trans. On Antennas Propag, 1986, 34(3): 276-280.). Hou and Wu first proposed that the time delay estimation problem can be transformed into the problem of sine frequency estimation. The sine frequency estimation model is equivalent to the DOA estimation model, so the MUSIC algorithm used for DOA estimation is applicable to time delay estimation, which effectively improves the resolution of multipath time delay estimation (Z Q Hou, Z D Wu. A new method for resolution estimation of time delay[J]. in Proc. ICASSP, 1982: 420-423.). But this method is not good for narrowband signal and signal with slowly changing envelope characteristics due to the inclusion of spectral division operation. Ge used the correlation results of the measured data to construct the covariance matrix, which realized the super-resolution time delay estimation of the signal with slowly changing envelope characteristics, but the estimation spectrum constructed by this method is still the traditional MUSIC spectrum, which has the problems of DP ambiguity and not steep enough spectrum peak (Ge F X, H Zhang, Yang J, et al. Super-resolution time delay estimation in multipath environments[J]. IEEE Trans. On Circuits and Systems I, 2007, 54(9): 1977-1986.). Li X et al. improved the covariance matrix by using diagonal loading method, which improved the steepness of the estimation spectrum. But this method is complex and has poor real-time performance (Li X, Ma X, Yan S Hou C. Super-resolution time delay estimation for narrowband signal[J]. IET Radar Sonar and Navigation, 2012, 6(8): 781-787.).

[0004] The spatial smoothing technology is used to improve the covariance matrix, which can reduce the dimension of matrix calculation and make the spectral peak more steep. The SSMUSIC algorithm is improved, the spatial smoothing idea is used to construct the covariance matrix, and the SSMUSIC algorithm idea is used to weight the estimated spectrum, so that the estimation accuracy of narrow-band signals and spectrum uneven signals is improved (Wang Yunlong, Wu Ying. Improved SSMUSIC super-resolution multipath time delay estimation algorithm [J]. Signal processing, 2014, 30(8): 979-986.). But the forward covariance matrix constructed by this method generally cannot meet the Hermitian matrix under limited observation data. Lu Jun et al. use the signal eigenvalue and noise power to construct the weighting value to correct the estimated spectrum, compared with the traditional MUSIC algorithm which only uses noise subspace information, it has higher time delay estimation accuracy and time delay resolution under the condition of limited data length and low signal-to-noise ratio (Lu Jun, Zhang Qunfei, Shi Wenta, et al. Subspace weighted multiple signal classification time delay estimation algorithm [J]. Journal of Detection and Control, 2021, 43(01): 30-35+40.).

[0005] In summary, the existing problems of the prior art are that the covariance matrix constructed under limited observation data has errors, and the time delay estimation performance is poor under low signal-to-noise ratio. SUMMARY

[0006] The purpose of the present application is to provide a multipath time delay estimation method based on improved MUSIC algorithm, so as to estimate the multipath time delay with high precision under the condition of small sampling data and low signal-to-noise ratio.

[0007] Technical scheme: the multipath time delay estimation method based on improved MUSIC algorithm, comprising the following steps:

[0008] (1) calculating the normalized cross spectrum sequence h[k] of the received signal in the multipath environment;

[0009] Under the multipath environment, the received signals of two receivers are represented as:

[0010]

[0011] In the formula, y1(n) and y2(n) represent the signals received by two receivers respectively, s(n) represents an unknown source signal, w1(n) and w2(n) represent additive white Gaussian noise, s(n), w1(n) and w2(n) are mutually independent; it is assumed that y1(n) has only direct wave, y2(n) contains multipath signal; D2 is the number of multipath propagation paths, α 2i represents the attenuation coefficient of each path, τ 2i represents the relative time delay between the multipath component of y2(n) and the direct wave component of y1(n).

[0012] The autocorrelation of the received signal y1(n) is:

[0013]

[0014] where r ss (m) and respectively represent the autocorrelation of s(n) and w1(n); the power spectrum of y1(n) is

[0015]

[0016] where S ss (w) and respectively represent the Fourier transform of r ss (m) and ;

[0017] The cross-correlation of the received signals y1(n) and y2(n) is:

[0018]

[0019] The cross-power spectrum of the received signals y1(n) and y2(n) is:

[0020]

[0021] Substituting equation (3) into equation (5) gives:

[0022]

[0023] Normalizing by gives the normalized cross-spectrum of the received signals:

[0024]

[0025] where

[0026] Sampling h(w) in the frequency domain gives the normalized cross-spectrum sequence h[k] of the received signals:

[0027]

[0028] The vector form of equation (8) is:

[0029] h = Aα - ε (9)

[0030] where h = [h[0], h[1],..., h[K-1]] T , ε = [ε[0], ε[1],..., ε[K-1]] T ,

[0031] (2) Divide h[k] into several overlapping sub-sequences h q , and take the conjugate of each sub-sequence data x q ;

[0032] Divide the data sequence h[k] into overlapping sub-sequences of length M, and take

[0033] h q = [h[q], h[q+1],..., h[q+M-1]] T q = 0, 1,..., K-M (10)

[0034] where the length of the sub-sequence is denotes the floor function; take the conjugate of each sub-sequence data h q Define a new data vector x q :

[0035]

[0036] where J is an M x M permutation matrix with 1 on the anti-diagonal and 0 elsewhere;

[0037] (3) Calculate the improved covariance matrix R q using h q and x

[0038] Calculate the mean of the covariance matrix of all sub-sequences h q , which is expressed as follows:

[0039]

[0040] Calculate the mean of the autocorrelation matrix of x q and the mean of the cross-correlation matrix of h q and x q

[0041]

[0042]

[0043] Define the improved covariance matrix R as:

[0044]

[0045] (4) Calculate the mean of the covariance matrix R ​​​Eigenvalue decomposition is performed to obtain noise eigenvalues and noise eigenvectors corresponding to the noise eigenvalues;

[0046] Eigenvalue decomposition is performed on the covariance matrix Eigenvalue decomposition is performed to obtain noise eigenvalues and noise eigenvectors corresponding to the noise eigenvalues; The eigenvalues after ordering represent the eigenvalues of the signal, the first D2 largest eigenvalues are the signal eigenvalues, and the remaining M-D2 eigenvalues are the noise eigenvalues, wherein the multipath number D2 is estimated by the minimum description length criterion, and then

[0047]

[0048] In the formula, ∑ S represents a diagonal matrix composed of signal eigenvalues, represents a signal subspace composed of signal eigenvectors corresponding to ∑ S , ∑ N represents a diagonal matrix composed of noise eigenvalues, represents a noise subspace composed of noise eigenvectors corresponding to ∑ N ;

[0049] (5) The noise eigenvalues are modified, and the modified noise eigenvalues are used to improve the MUSIC time delay estimation spectrum;

[0050] The M-D2 noise eigenvalues are modified:

[0051]

[0052] In the formula, is the modified noise eigenvalue, β is the modification value, which is used to control the divergence degree between noise eigenvalues; β is selected as the minimum integer satisfying formula (18);

[0053]

[0054] The orthogonality of the signal subspace and the noise subspace is used to construct the MUSIC time delay estimation spectrum, and the modified noise eigenvalues are used to weight the estimation spectrum, and the expression of the new MUSIC time delay estimation spectrum is:

[0055]

[0056] In the formula, v i is the eigenvector corresponding to the noise eigenvalue λ i ;

[0057] (6) The estimation spectrum is searched for a spectral peak to obtain a multipath time delay value;

[0058] The spectral peak search is performed on formula (19), and the time τ corresponding to the first D2 largest peak values obtained is the estimated multipath time delay value.

[0059] A computer storage medium storing a computer program that, when executed by a processor, implements the aforementioned multipath delay estimation method based on the improved MUSIC algorithm.

[0060] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multipath delay estimation method based on the improved MUSIC algorithm described above.

[0061] Beneficial Effects: Compared with existing technologies, the present invention has the following advantages: The method provided by the present invention can make the spectral peaks of the MUSIC time delay estimation spectrum steeper, and has better time delay resolution under conditions of low signal-to-noise ratio and inaccurate path number estimation. The time delay estimation performance of the present invention is superior to the traditional MUSIC algorithm and the improved SSMUSIC algorithm under different signal-to-noise ratios and different numbers of computation points. Attached Figure Description

[0062] Figure 1 Here is a flowchart of a multipath delay estimation method based on the improved MUSIC algorithm;

[0063] Figure 2 This is a comparison chart of the covariance matrix before and after improvement; where, Figure 2 (a) shows the covariance matrix before improvement, and Figure 2(b) shows the covariance matrix after improvement;

[0064] Figure 3 This is a comparison chart of the MUSIC time delay estimation spectrum before and after weighting; where, Figure 3 (a) is the MUSIC time delay estimation spectrum before weighting. Figure 3 (b) is the weighted spectrum of the MUSIC time delay estimation;

[0065] Figure 4 The simulation results are shown in the figure under different signal-to-noise ratio conditions;

[0066] Figure 5 The simulation results are shown in the figure under different data point conditions. Detailed Implementation

[0067] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0068] like Figure 1 As shown, a multipath delay estimation method based on an improved MUSIC algorithm includes the following steps:

[0069] Step 1: Calculate the normalized cross-spectral sequence h[k] of the received signal under multipath conditions.

[0070] In a multipath environment, the received signals from the two receivers, after sampling, are represented as follows:

[0071]

[0072] In the formula, y1(n) and y2(n) represent the signals received by the two receivers respectively, s(n) represents the unknown source signal, and w1(n) and w2(n) represent additive white Gaussian noise. s(n), w1(n), and w2(n) are uncorrelated. It is assumed that y1(n) contains only direct waves, and y2(n) contains multipath signals. D2 is the number of multipath propagation paths, and α... 2i τ represents the attenuation coefficient of each path. 2i This represents the relative time delay between the multipath component of y2(n) and the direct wave component of y1(n).

[0073] The autocorrelation of the received signal y1(n) is:

[0074]

[0075] In the formula, r ss (m) and Let represent the autocorrelation of s(n) and w1(n), respectively. The power spectrum of y1(n) is...

[0076]

[0077] In the formula, S ss (w) and They represent r respectively ss (m) and Fourier transform.

[0078] The cross-correlation between the received signals y1(n) and y2(n) is as follows:

[0079]

[0080] The cross-power spectrum of the received signals y1(n) and y2(n) is as follows:

[0081]

[0082] Substituting equation (3) into equation (5), we get:

[0083]

[0084] use right Normalization is performed to obtain the normalized cross-spectrum of the received signal:

[0085]

[0086] In the formula,

[0087] The normalized cross-spectrum sequence h[k] of the received signal is obtained by frequency domain sampling of h(w):

[0088]

[0089] The vector form of equation (8) is:

[0090] h = Aα - ε (9)

[0091] where h = [h[0], h[1],..., h[K-1]] T ε = [ε[0], ε[1],..., ε[K-1]] T ,

[0092] Taking h[k] as the array received data in DOA estimation, equation (8) is consistent with the DOA estimation model under the uniform linear array (ULA) structure, so the MUSIC algorithm in DOA estimation can be applied to multipath time delay estimation.

[0093] Step two, divide h[k] into multiple overlapping subsequences h q using the spatial smoothing idea, and take the conjugate data x q of each subsequence.

[0094] Divide the data sequence h[k] into overlapping subsequences of length M, and take

[0095] h q = [h[q], h[q+1],..., h[q+M-1]] T q = 0, 1,..., K-M (10)

[0096] where the length of the subsequence represents the floor function. Take the conjugate form of each subsequence data h q Define a new data vector x q :

[0097]

[0098] where J is an M x M permutation matrix with 1 on the anti-diagonal and 0 elsewhere.

[0099] Step three, calculate the improved covariance matrix R q using h q and x

[0100] Calculate all the subsequences h​q The mean of the covariance matrix is ​​expressed as follows:

[0101]

[0102] Calculate x q The mean of the autocorrelation matrix and h q With x q The mean of the cross-correlation matrix

[0103]

[0104]

[0105] Define the improved covariance matrix for:

[0106]

[0107] Improved covariance matrix It effectively utilizes the conjugate information of the measurement data to meet the requirements. Its elements are conjugate symmetric on the main diagonal, which further satisfies the properties of the Hermitian matrix.

[0108] Step four, for Perform eigenvalue decomposition to obtain noise eigenvalues ​​and their corresponding noise eigenvectors.

[0109] For covariance matrix Perform eigenvalue decomposition, in order to The features are represented by the sorted values. The first D2 largest features are signal features, and the remaining M-D2 features are noise features. The number of multipath features, D2, is estimated by the minimum description length criterion.

[0110]

[0111] In the formula, ∑ S This represents a diagonal matrix composed of the signal's eigenvalues. This indicates that it corresponds to ∑ S The signal subspace composed of signal feature vectors, ∑ N This represents a diagonal matrix composed of noise eigenvalues. This indicates that it corresponds to ∑ N The noise subspace is composed of noise feature vectors.

[0112] Step 5: Correct the noise characteristic values ​​and use the corrected noise characteristic values ​​to improve the MUSIC time delay estimation spectrum.

[0113] The M-D2 noise eigenvalues are corrected:

[0114]

[0115] In the formula, is the corrected noise eigenvalue, and β is a correction value for controlling the divergence degree between the noise eigenvalues. The selection of the correction value β directly affects the performance of the algorithm. If β is too large, the difference between the weights is small, and the algorithm performance is close to that of the traditional equal weight matrix, and the algorithm performance cannot be improved. If β is too small, the noise eigenvalues diverge at a high degree under a low signal-to-noise ratio, and the algorithm performance is unstable. According to the information theory criterion, under the condition that the number of the signal sources can be correctly estimated, the ratio of the maximum value to the minimum value of the noise eigenvalues is not greater than 2, and the corrected noise eigenvalue is substituted into formula (18), that is, formula (18), and the minimum integer satisfying formula (18) is selected as β of the present application.

[0116]

[0117] The orthogonalities of the signal subspace and the noise subspace are utilized to construct the MUSIC time delay estimation spectrum, and the new MUSIC time delay estimation spectrum is obtained by weighting the estimation spectrum by using the corrected noise eigenvalue, and the expression of the new MUSIC time delay estimation spectrum is as follows:

[0118]

[0119] In the formula, v i is the eigenvector corresponding to the noise eigenvalue λ i . When the signal-to-noise ratio is low, the noise eigenvalues diverge, and the difference between the signal eigenvalues is small. The corrected noise eigenvalue is used to change the weight of the projection component of a(τ) in the noise subspace, the difference between the noise eigenvalues is balanced, and the noise interference can be reduced to a certain extent.

[0120] Step six, spectrum peak searching is performed on the estimation spectrum to obtain the multipath time delay value.

[0121] Since the signal subspace and the noise subspace are orthogonal, the vector a(τ 2i ) falls in the signal subspace, and In the actual propagation environment, there is noise, so that a(τ 2i ) and are not completely orthogonal, and therefore the time delay estimation needs to be realized by minimum optimization searching, that is, the spectrum peak searching is performed on formula (19), and the time τ corresponding to the first D2 maximum peak value is the estimated multipath time delay value.

[0122] The application effect of the present application will be described in detail in combination with simulation.

[0123] To evaluate the performance of the present application, simulation experiments are performed using a chirp signal as the transmitted signal. The expression of the chirp signal is as follows:

[0124]

[0125] where β = π · (f2-f1) / N, fi and f2 represent the lowest and highest frequencies of the signal respectively, α = 2π·fi, represents the random initial phase of the signal. The simulation is set to fi = 0.3, f2 = 0.5, and the defined bandwidth The number of multipaths D2 is set to 2, the signal time delay τ 21 = 5, τ 22 = 8, and the time delay difference between the two signals belongs to the super-resolution case. The number of data points N = 48, and the length of the normalized cross-spectrum sequence K = 95.

[0126] Simulation Experiment 1: Comparison before and after improving the covariance matrix

[0127] The simulation is set to α 21 = 1, α 22 = 0.8, and the signal-to-noise ratio is -8dB. The MUSIC time delay estimation spectrum images before and after improving the covariance matrix using the method of the present application are shown in Figure 2 Comparing Figure 2 (a) and Figure 2 (b), it can be seen that the peak of the improved MUSIC time delay estimation spectrum is steeper, the spectrum line at the non-time delay point is smoother, and the time delay estimation accuracy is higher.

[0128] Simulation Experiment 2: Comparison before and after weighting the MUSIC time delay estimation spectrum

[0129] The simulation is set to α 21 = 0.25, α 22 = 1, and the signal-to-noise ratio is -11dB. At this time, the number of multipaths estimated using the minimum description length criterion is 1. The effects before and after weighting the MUSIC time delay estimation spectrum using the weight matrix designed in the present application are shown in Figure 3 Comparing Figure 3 (a) and Figure 3 (b), under the condition of inaccurate multipath estimation, the MUSIC time delay estimation spectrum image before weighting has only one peak value, and the MUSIC time delay estimation spectrum image after weighting has two peak values, which can distinguish the first path. It is shown that weighting can reduce noise interference to some extent, and the improved MUSIC algorithm has better time delay resolution.

[0130] Simulation Experiment 3: Performance comparison of three algorithms under different signal-to-noise ratios

[0131] In the case of data points N = 48, signal-to-noise ratio SNR = -10:2:4dB, the time delay estimation performance of the method of the application, the traditional MUSIC algorithm and the improved SSMUSIC algorithm are compared. The simulation is set to α 21 = 1, α 22 = 0.8, and 100 Monte Carlo simulation experiments are performed. The root mean square error (RMSE) of the three algorithms with respect to the change of the signal-to-noise ratio is shown in Figure 4 It can be seen from Figure 4 that the time delay estimation performance of each algorithm improves as the signal-to-noise ratio improves. Compared with the method of the application and the improved SSMUSIC algorithm which adopts spatial smoothing preprocessing technology, the traditional MUSIC algorithm has poorer performance. Under different signal-to-noise ratios, the RMSE of the method of the application is smaller than that of the improved SSMUSIC algorithm, and when the signal-to-noise ratio SNR≤-4dB, the RMSE of the method of the application is 0.01 or more smaller than that of the improved SSMUSIC algorithm. Therefore, the time delay estimation performance of the method of the application is superior to that of the traditional MUSIC algorithm and the improved SSMUSIC algorithm.

[0132] Simulation experiment 4: Performance comparison of three algorithms under different calculation points

[0133] In the case of signal-to-noise ratio SNR = -6dB, data points N = 20:10:100, the time delay estimation performance of the method of the application, the traditional MUSIC algorithm and the improved SSMUSIC algorithm are compared. The simulation is set to α 21 = 1, α 22 = 0.8, and 100 Monte Carlo simulation experiments are performed. The RMSE of the three algorithms with respect to the change of the data points is shown in Figure 5 It can be seen from Figure 5 that as the number of data points increases, the RMSE of the three algorithms decreases. The RMSE of the method of the application is also smaller than that of the traditional MUSIC algorithm and the improved SSMUSIC algorithm when the number of data points is small, and when the number of data points N≥30, the RMSE can be controlled within 0.1, indicating that the method of the application has more advantages under small sample conditions.

[0134] According to the above experimental results, the following conclusions are obtained:

[0135] The method of the application has a steeper spectrum peak, has better time delay resolution under the condition of low signal-to-noise ratio and inaccurate range estimation, and has better time delay estimation performance than the traditional MUSIC algorithm and the improved SSMUSIC algorithm under different signal-to-noise ratios and different calculation points.

Claims

1. A method for multipath time delay estimation based on an improved MUSIC algorithm, characterized in that, The method comprises the following steps: (1) calculating a normalized cross-spectrum sequence h[k] of the received signal in a multipath environment; The received signals of two receivers in a multipath environment are represented as: In the formula, y1(n) and y2(n) represent signals received by two receivers respectively, s(n) represents an unknown source signal, w1(n) and w2(n) represent additive white Gaussian noises, s(n), w1(n) and w2(n) are mutually irrelevant; it is assumed that y1(n) only has a direct wave, and y2(n) contains multi-path signals; D2 is the number of multi-path propagation paths, α 2i represents an attenuation coefficient of each path, τ 2i represents a relative time delay between a multi-path component of y2(n) and a direct wave component of y1(n); The autocorrelation of the received signal y1(n) is: where r ss (m) and respectively, and the power spectrum of y1(n) is wherein S ss (w) and represent the Fourier transform of r ss (m) and respectively. The cross-correlation of the received signals y1(n) and y2(n) is: The cross-power spectrum of the received signals y1(n) and y2(n) is: Substitute equation (3) into equation (5) to obtain: With To The normalized received signal cross-spectrum is obtained by normalization: In the formulae, The normalized cross-spectrum sequence h[k] of the received signal is obtained by frequency domain sampling of h(w): The vector form of equation (8) is: h=Aα-ε (9) In the formula, h = [h[0], h[1],..., h[K-1]] T , ε = [ε[0], ε[1],..., ε[K-1]] T , (2) Using the idea of spatial smoothing, the h[k] is divided into multiple overlapping sub-sequences h q , and the conjugate data x q of each sub-sequence is taken. The data sequence h[k] is divided into overlapping sub-sequences with a length of M, and h[k] is taken as h q = [h [q], h [q + 1],..., h [q + M - 1]] T q = 0, 1,..., K - M (10) In the formula, the length of the subsequence This indicates rounding down; it takes the data h from each subsequence. q conjugate form Define a new data vector x q : In the formula, J is an M×M order exchange matrix with 1 on the anti-diagonal line and 0 on the rest; (3) using h q and x q The improved covariance matrix is calculated Compute the mean of the covariance matrix of all sub-sequences h q whose expression is as follows: Compute x q Mean of autocorrelation matrix of x and h q Mean of cross-correlation matrix of x q and h Defining an improved covariance matrix is: (4) to perform eigenvalue decomposition to obtain noise eigenvalues and noise eigenvectors corresponding thereto; Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues representing the sorted eigenvalues, the first D2 largest eigenvalues are signal eigenvalues, and the remaining M-D2 eigenvalues are noise eigenvalues, where the number of multipaths D2 is estimated by the minimum description length criterion, and further In the formula, ∑ S This represents a diagonal matrix composed of the signal's eigenvalues. This indicates that it corresponds to ∑ S The signal subspace composed of signal feature vectors, ∑ N This represents a diagonal matrix composed of noise eigenvalues. This indicates that it corresponds to ∑ N The noise subspace composed of noise feature vectors; (5) correcting the noise eigenvalues, and improving the MUSIC time delay estimation spectrum by using the corrected noise eigenvalues; The M-D2 noise eigenvalues are corrected: In the formula, is the corrected noise characteristic value, β is a correction value for controlling the divergence degree between noise characteristic values; β is selected as the minimum integer satisfying formula (18); The orthogonality of the signal subspace and the noise subspace is used to construct the MUSIC time delay estimation spectrum, and the corrected noise eigenvalues are used to weight the estimation spectrum, so that the expression of the new MUSIC time delay estimation spectrum is: In the formula, v i is an eigenvector corresponding to the noise eigenvalue λ i . (6) performing spectrum peak searching on the estimation spectrum to obtain the multipath time delay value; The time τ corresponding to the first D2 maximum peak values obtained by performing spectrum peak searching on equation (19) is the estimated multipath time delay value.

2. A computer storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the multipath time delay estimation method based on the improved MUSIC algorithm in claim 1.

3. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the multipath time delay estimation method based on the improved MUSIC algorithm in claim 1.

Images