Radar multipath signal parameter estimation method in closed environment

Abstract

The invention discloses a radar multipath signal parameter estimation method in a closed environment, belongs to a signal processing technology, and particularly relates to a sparse reconstruction technology. According to the method, an over-complete dictionary is designed according to transmitted waveform parameters, resolution and precision requirements, and then multi-path signal parameter estimation is realized by utilizing multi-path space sparsity and combining real part and imaginary part characteristics of beat signals and the inner product of the over-complete dictionary. The method has the advantages that compared with a traditional CS method, the precision and robustness of radar multipath signal parameter estimation in the closed environment are effectively improved, and interference of the multipath effect on radar signals in the closed environment is reduced; compared with a traditional CS method, the method has higher global convergence and better anti-noise performance, and has higher reconstruction precision and success rate in a low signal-to-noise ratio environment; the over-complete dictionary of the system can be adaptively adjusted according to different requirements on resolution and operand so as to meet different requirements, and the operation complexity is low.

Description

Technical Field

[0001] The present invention belongs to the field of signal processing, and particularly relates to sparse reconstruction technology. Background Art

[0002] Radar is widely used in closed environments, such as elevator ranging, train anti-collision and obstacle detection in tunnels, underground rail transit, indoor positioning, in-vehicle personnel detection, etc. Affected by the wall boundaries, signals are subject to multipath interference during propagation in closed environments, resulting in signal fading and the phenomenon of "false targets".

[0003] The multipath effect has a great impact on applications that require high-precision ranging by radar, such as elevator ranging and indoor positioning. The key to improving the radar ranging accuracy in closed environments lies in reducing the interference of the multipath effect on target echo detection in closed environments. By accurately estimating the parameters of multipath components, the direct wave and multipath can be effectively distinguished, thereby reducing the interference. Therefore, the present invention studies the method for estimating the parameters of radar multipath signals in closed environments to solve the multipath effect problem.

[0004] Representative methods for estimating the parameters of multipath signals include the cyclic correlation method, the Expectation Maximization (EM) method, the method based on Weighted Fourier transform and Relaxation (WRELAX), and the method based on Compressed Sensing (CS). Although the cyclic correlation method is effective, it has great limitations in terms of parameter estimation accuracy. The EM method reduces the computational complexity through dimensionality reduction, but requires prior estimation. The WRELAX method solves the initial value problem, but has a high computational complexity and is prone to falling into local optima. The method based on CS provides higher resolution compared to the above traditional methods and has been widely applied. The specific implementation of the CS method is to first mix the echo signal with the transmitted signal to obtain a beat signal, then sample the beat signal with an Analog-Digital Converter (ADC), utilize the spatial sparsity of the beat signal to construct an overcomplete dictionary in the digital domain, and according to the compressed sensing theory, calculate the correlation degree between the column vectors of the overcomplete dictionary and the beat signal and compare to select the larger atoms in each iteration through a greedy method, thereby realizing the estimation of the amplitude and delay of the multipath signal.

[0005] The sparse reconstruction theory indicates that if a signal exhibits sparse characteristics in a certain transform domain, it can be sampled at a rate much lower than that required by the Nyquist sampling theorem and still achieve high-precision reconstruction. According to the analysis of the propagation model of multipath signals in a closed environment, the number of effective reflection paths with strong energy in space and that can be received by the radar is much smaller than the total number of all reflection paths. Therefore, the received signal has obvious sparsity in space, enabling approximate reconstruction of the signal with only a small number of measurement values. The core of the method for estimating multipath component parameters based on sparse decomposition includes the sparse representation of the signal and its accurate reconstruction.

[0006] Through the CS method, high resolution can be achieved, but it requires a high signal-to-noise ratio (SNR), has high computational complexity, and poor global convergence. Summary of the Invention

[0007] The technical problem to be solved by the present invention is that the traditional CS method faces the problem of the degradation of the Restricted Isometry Property (RIP) in practical applications. The fundamental reason is that there is a strong correlation between different columns in the overcomplete dictionary adopted by the CS method, which weakens the stability and accuracy of sparse signal reconstruction. Due to the degradation of the RIP property, the CS method may result in higher errors when reconstructing sparse signals, affecting the accuracy and robustness of parameter estimation, especially more significantly in low SNR or high-dimensional complex environments. To solve this problem, a method for estimating radar multipath signal parameters based on sparse reconstruction combined with complex signal characteristics in a closed environment is now proposed.

[0008] To break the limitation of the strong correlation between different column vectors of the overcomplete dictionary on the reconstruction accuracy, the present invention analyzes the real and imaginary part characteristics of the inner product between the beat signal and the overcomplete dictionary by utilizing the spatial sparse characteristics of the multipath signal, realizes the effective identification of the direct wave and multipath components, thereby improving the accuracy and robustness of multipath signal parameter estimation. Accordingly, a method for estimating radar multipath signal parameters in a closed environment is proposed, and this method includes:

[0009] Step 1: Design an overcomplete dictionary according to the parameters of the transmitted signal and the resolution requirements for parameter estimation;

[0010] Step 2: Calculate the inner product w of Ψ H ·x, where Ψ H and the beat signal x = [x(1), x(2), …, x(N)] T , where Ψ H is the Hermitian transformation of Ψ, and w is an M×1-dimensional column vector used to represent the result of path matching;

[0011] Affected by the multipath effect, the beat signal x is expressed as the direct wave component x d =[x d (1),x d (2),…,x d (N)] T plus the multipath component x m =[x m (1),x m (2),…,x m (N)] T where x d (1),x d (2),…,x d (N) represents the direct wave component at each sampling point, and x m (1),x m (2),…,x m (N) represents the multipath component at each sampling point; there are components corresponding to the direct wave and multipath components in the column vector w where w n represents the path matching result of the direct wave component, and w m represents the path matching result of the multipath component, and Ψ :,n and Ψ :,m represent the nth column and the mth column of the overcomplete dictionary Ψ respectively; expand the direct wave component w n and the multipath component w m ;

[0012] According to the imaginary part property Im(w n )·Im(w m )<0, screen out all element pairs that satisfy the imaginary part property from the column vector w, and the column vectors corresponding to them in the overcomplete dictionary form:

[0013] Ψ pi =[Ψ :,n ,Ψ :,m ∈Ψ pair

[0014] where Ψ pair =[Ψ p1 ,Ψ p2 ,…,Ψ pt ,…,Ψ pi represents the path pair matrix;

[0015] Step 3: Calculate the amplitude ratio k n of the path pairs in Ψ m using the imaginary parts of w pi ∈Ψ pair , and according to k pair , pairSelect the path pairs Ψ n that satisfy |Im(w m )| < |Im(w pi )|, and update Ψ pair ;

[0016] Step 4: Calculate the amplitude pairs a n , a m corresponding to the direct wave and multipath components according to the linear equations formed by the real parts of w pair and k pi = (a 1n , a 2m ) T ∈ A pair , where a 1n , a 2m represent the calculated amplitudes of the direct wave and multipath components. The amplitude pair matrix A pair = [a p1 , a p2 , …, a pt , …, a pi ;

[0017] Step 5: Select the Ψ pt and a pt that meet the requirements to reconstruct x' pt , where x' pt represents the reconstructed signal of the beat signal x;

[0018] Calculate the residual χ = ||x - x' pt || < δ, where δ represents the set residual value. Select the x' pt that meets the residual requirement, and output the corresponding Ψ pt ∈ Ψ pair , a pt ∈ A pair as the corresponding parameters for radar multipath signal estimation.

[0019] Furthermore, the specific method of step 1 is as follows:

[0020] Assume that the center frequency of the transmitted signal is f 0 , the speed of light is c, the wavelength of the transmitted signal is , the bandwidth is B, the frequency sweep time is T c , the frequency modulation slope is K, and the number of sampling points is N. Then the digitized beat signal x(n) is expressed as:

[0021]

[0022] where I represents the number of reflection paths, a i and R i respectively represent the amplitude and distance corresponding to the i-th path, and nΔ represents the n-th sampling point;

[0023] Preprocess the signal to determine the range cell containing the direct wave and multipath components, denoted as R b ; Select the grid ratio G of the overcomplete dictionary r , and the dictionary interval ΔR' = ΔR / G r , where ΔR represents the range resolution of the radar; the range information corresponding to different column vectors in the overcomplete dictionary is expressed as l = [l 1 , l 2 , …, l M T , where M represents the number of columns of the dictionary; the elements in I are spaced by ΔR', and its range is expressed as where l min , l max represent the minimum and maximum elements in I;

[0024] Construct an N×M-dimensional overcomplete dictionary Ψ according to the beat signal x(n) and the range information vector l. The element in the nth row and mth dimension of Ψ where nΔ represents the nth sampling point, and l m ∈l.

[0025] Furthermore, the specific method for expanding the direct wave component w n and the multipath component w m in terms of real and imaginary parts in step 2 is as follows:

[0026]

[0027] where a 1 , a 2 correspond to the amplitudes of the direct wave and multipath components respectively, and R 1 , R 2 represent the ranges of the direct wave and multipath components respectively.

[0028] The block diagram of the radar multipath signal parameter estimation method based on sparse reconstruction combined with complex signal characteristics is as Figure 1 shown.

[0029] The beneficial effects of the present invention are:

[0030] Compared with the traditional CS method, this method effectively improves the accuracy and robustness of radar multipath signal parameter estimation in a closed environment, and reduces the interference of multipath effects on radar signals in a closed environment;

[0031] Compared with the traditional CS method, this method has stronger global convergence and better anti-noise performance, and has higher reconstruction accuracy and success rate in a low signal-to-noise ratio environment;

[0032] ​The over-complete dictionary of this system can be adaptively adjusted according to different requirements for resolution and computational complexity to meet different needs, and the computational complexity is low. Description of the Drawings

[0033] Figure 1 It is a block diagram of a radar multipath signal parameter estimation method based on sparse reconstruction combined with complex signal characteristics.

[0034] Figure 2 It is the reconstruction results of two parameter estimation methods in a closed multipath environment; (a) is the result diagram of this method, and (b) is the result diagram of the traditional CS method.

[0035] Figure 3 It is the mean absolute error curves of two parameter estimation methods in a closed multipath environment; (a) is the curve diagram of this method, and (b) is the curve diagram of the traditional CS method.

[0036] Figure 4 It is the reconstruction success rate curves of two parameter estimation methods in a closed multipath environment; (a) is the curve diagram of this method, and (b) is the curve diagram of the traditional CS method. Detailed Implementation Manner

[0037] Step 1: Use Linear Frequency Modulated Continuous Wave (LFMCW) as the transmitted waveform, and let the center frequency f of the transmitted LFMCW signal 0 = 77 GHz, the bandwidth B = 500 MHz, and the frequency modulation slope K = 1.56×10 13 . Simulate the multipath effect in a closed environment. The range resolution of the radar ΔR = 0.3 m. The distances from the direct wave and the reflection points corresponding to the multipath components are 3 m and 3.06 m respectively, and the normalized amplitudes of the signals are 1 and 0.5 respectively. Let the sampling rate f s of the ADC be 4 MHz, and the number of sampling points N be 1024. Simulate and generate a digital beat baseband signal.

[0038] Step 2: Select the signal-to-noise ratio to be 10 dB, and the grid ratios are all set to 10, and use the digital beat baseband signal generated in Step 1. The method in this paper first pre-estimates the beat signal, then designs an over-complete dictionary using the digital beat baseband signal generated in Step 1, and uses the real and imaginary characteristics after the inner product of it and the beat signal to estimate the distance and amplitude information of the direct wave and the multipath components. The measurement results are as Figure 2 (a) shown. Subsequently, use the traditional CS method to process the same signal, reconstruct the direct wave and the multipath components, and the results are as Figure 2 (b) shown. Compare the parameter estimation results of the traditional CS method and the method described in this paper in a closed multipath environment. As Figure 2 (a) andFigure 2 As shown in Fig. (b), the direct wave and multipath components reconstructed by the traditional CS method are located at 2.94 m and 3.00 m respectively, and the amplitudes are 0.3942 and 0.4592 respectively, which do not match the parameters set in the simulation. This indicates that the traditional CS method fails to reconstruct and the parameter estimation is inaccurate under the conditions of a signal-to-noise ratio of 10 dB and a grid ratio of 10. For the method proposed in this paper, the direct wave and multipath components are located at 3.00 m and 3.06 m respectively, and the amplitudes are 0.9919 and 0.4862 respectively, enabling accurate estimation of the distance and amplitude, and realizing the parameter estimation of radar multipath signals under the conditions of low signal-to-noise ratio and high grid ratio.

[0039] Step 3: To more intuitively evaluate the accuracy of radar multipath signal parameter estimation, the Mean Absolute Error (MAE) is defined as the mean absolute error between the reconstructed signal x' and the beat signal x. Select grid ratios of 2, 4, 6, and 8 respectively, and the signal-to-noise ratio steps from -20 dB to 20 dB in 2 dB increments as the experimental conditions. Under each condition, 100 Monte Carlo experiments are conducted for each method. First, use the method proposed in this paper to estimate the parameters of the digital beat baseband signal generated in Step 1 to obtain the MAE results, as shown in Figure 3 Fig. (a). Subsequently, use the traditional CS method for parameter estimation to obtain the MAE results, as shown in Figure 3 Fig. (b). Comparing Figure 3 Figs. (a) and (b), it can be seen that as the signal-to-noise ratio increases, the MAE of both methods shows a downward trend, and the MAE of the method proposed in this paper is significantly lower overall. At the same SNR, when attempting to increase the grid ratio, the MAE of the CS method continues to increase, which is caused by the degradation of the RIP property of the overcomplete dictionary. In contrast, the MAE of the method proposed in this paper decreases significantly as the grid ratio increases. This is attributed to the higher convergence of the method proposed in this paper, and at the same time, a larger grid ratio further improves the estimation accuracy. Table 1 summarizes and compares the performance of different indicators of the traditional CS method and the method proposed in this paper at a signal-to-noise ratio of 10 dB. As shown in Table 1, at a signal-to-noise ratio of 10 dB, when the grid ratio is 2, the mean absolute error of the method proposed in this paper is 2.622, while that of the CS method is 3.597, and the two are relatively close. However, when the grid ratio is increased to 8, the MAE of the method proposed in this paper drops to 1.365, while the MAE of the CS method rises to 13.029, which fully demonstrates the significant advantage of the method proposed in this paper in terms of parameter estimation accuracy.

[0040] Step 4: Select grid ratios of 2, 4, 6, and 8 respectively, and the signal-to-noise ratio ranges from -20 dB to 20 dB with a step of 2 dB. Under each condition, 100 Monte Carlo experiments are carried out for each of the two methods. First, using the digital beat baseband signal generated in Step 1, parameter estimation is performed by the method proposed in this paper, and the reconstruction success rate is as shown in Figure 4 (a); Subsequently, the traditional CS method is used for parameter estimation, and the reconstruction success rate is as shown in Figure 4 (b). Compare the reconstruction success rates of the traditional CS method and the method proposed in this paper under different signal-to-noise ratios and grid ratios, as shown in Figure 4 (a) and Figure 4 (b). Generally speaking, under low signal-to-noise ratio conditions, the reconstruction rates of both methods are relatively low, and they gradually increase with the increase of SNR, and finally tend to be stable at a relatively high reconstruction rate level. Among them, the method proposed in this paper is always superior to the traditional CS method in terms of reconstruction rate. It is particularly worth noting that even when the grid ratio is increased, the method proposed in this paper can still maintain a high reconstruction rate, while under the same conditions, the reconstruction rate of the CS method drops significantly. As shown in Table 1, when the signal-to-noise ratio is 10 dB and the grid ratio is 2, the reconstruction rates of both methods reach 1.00; however, when the grid ratio is increased to 8, the reconstruction rate of the method proposed in this paper still remains at 0.98, which is significantly higher than 0.70 of the CS method. This fully demonstrates that the method proposed in this paper has significant advantages in terms of the robustness and effectiveness of parameter estimation.

[0041] Through experimental verification, this method effectively improves the accuracy and robustness of radar multipath signal parameter estimation in a closed environment, and its average absolute error, reconstruction success rate, and operation time of parameter estimation are all superior to the traditional CS method.

[0042] Table 1 shows the performance indicators of the two parameter estimation methods when the signal-to-noise ratio is 10 dB;

[0043]

Claims

1. A method for estimating radar multipath signal parameters in a closed environment, the method comprising: Step 1: Design an over-complete dictionary based on the parameters of the transmitted signal and the resolution requirements for parameter estimation; Step 2: According to w = Ψ H x, calculate Ψ H And the beat signal x=[x(1),x(2),…,x(N)] T The inner product w of H is the Hermitian change of Ψ, w is an M×1-dimensional column vector, used to represent the result of path matching; Affected by the multipath effect, the beat signal x is expressed as the direct wave component x d =[x d (1),x d (2),…,x d (N)] T and the multipath component x m =[x m (1),x m (2),…,x m (N)] T The accumulation of x d (1),x d (2),…,x d (N) represents the direct wave component of each sampling point, x m (1),x m (2),…,x m (N) represents the multipath component of each sampling point; the column vector w contains components corresponding to the direct wave and the multipath component where w n represents the path matching result of the direct wave component, w m represents the path matching result of the multipath component, Ψ :,n With :,m Respectively represent the nth and mth columns of the overcomplete dictionary Ψ; expand the direct wave component w according to the real and imaginary parts n and multipath component w m ; in accordance with The imaginary part characteristic Im(w n )·Im(w m )<0, all element pairs that satisfy the imaginary part characteristics are selected from the column vector w, and the corresponding column vectors in the overcomplete dictionary are composed of: P pi =[Ψ :,n ,P :,m ]∈Ψ pair Among them Ψ pair =[Ψ p1 ,Ψ p2 ,…,Ψ pt ,…,Ψ pi ] represents the path pair matrix; Step 3: Using w n , w m Calculate the imaginary part of Ψ pi ∈Ψ pair The amplitude ratio k of the path pair pair , according to k pair Filter out the ones that satisfy |Im(w n )|<|Im(w m )|Path pair Ψ pi , update pair ; Step 4: According to w n , w m The real part and k pair The linear equations formed calculate the amplitude pair a corresponding to the direct wave and the multipath component pi =(a 1n ,a 2m ) T ∈A pair , where a 1n ,a 2m A represents the calculated direct wave and multipath component amplitudes pair =[a p1 ,a p2 ,…,a pt ,…,a pi ] represents the magnitude pair matrix; Step 5: Select the one that meets the requirements pt and a pt Reconstruct x' pt , where x' pt represents the reconstructed signal of the beat signal x; Calculate the residual χ=||xx' pt ||<δ, δ represents the set residual value, select x' that meets the residual requirements pt , output the corresponding Ψ pt ∈Ψ pair , a pt ∈A pair As the corresponding parameter for radar multipath signal estimation.

2. The method for estimating radar multipath signal parameters in a closed environment as claimed in claim 1, characterized in that: The specific method of step 1 is: Assume that the center frequency of the transmitted signal is f0, the speed of light is c, and the wavelength of the transmitted signal is Bandwidth B, sweep time T c , the frequency modulation slope is K, the number of sampling points is N, then the digitally processed beat signal x(n) is expressed as: Where I represents the number of reflection paths, a i and R i They represent the amplitude and distance corresponding to the i-th path, respectively, and nΔ represents the n-th sampling point; Preprocess the signal to determine the distance unit containing the direct wave and multipath components, denoted as R b ; Select the grid ratio G of the overcomplete dictionary r , dictionary interval ΔR'=ΔR / G r , where ΔR represents the range resolution of the radar; the distance information corresponding to different column vectors in the overcomplete dictionary is expressed as l = [l1, l2, …, l M ] T , where M represents the number of columns in the dictionary; the elements in I are spaced by ΔR', and their range is expressed as Among them l min , l max Represents the smallest and largest elements in I; According to the beat signal x(n) and the distance information vector l, an N×M-dimensional overcomplete dictionary Ψ is constructed. The n-th row and m-th dimension element in Ψ Where nΔ represents the nth sampling point, l m ∈l.

3. The radar multipath signal parameter estimation method in a closed environment as claimed in claim 1, characterized in that: In step 2, the direct wave component w is expanded according to the real part and the imaginary part. n and multipath component w m The specific method is: in a1 and a2 correspond to the amplitudes of the direct wave and the multipath component, respectively, and R1 and R2 represent the distances of the direct wave and the multipath component, respectively.

Images