A method for eliminating automotive millimeter-wave radar interference signals based on sparse low-rank decomposition

Abstract

The present invention relates to a method for eliminating interference signals from automotive millimeter-wave radars based on sparse low-rank decomposition. This method considers the standard deviation of the interference signal and the radar signal during interference signal elimination, eliminating target signal loss and enabling the generation of a complete and accurate target signal. Furthermore, the method eliminates the need for modern algorithms to optimize the model, offering excellent performance and requiring no prior knowledge. Therefore, it can be implemented on automotive millimeter-wave radars, demonstrating its potential for application.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signal processing, and in particular to a method for eliminating automobile millimeter-wave radar interference signals based on sparse low-rank decomposition. Background Art

[0002] Automotive millimeter-wave radars offer all-weather capabilities and are crucial for autonomous driving. However, using the same spectrum can lead to mutual interference. According to International Telecommunication Union standards, automotive millimeter-wave radars on roads share the same spectrum, and autonomous vehicles typically deploy multiple radars around the vehicle. Therefore, with the increasing prevalence of autonomous driving and the increasing number of automotive millimeter-wave radars on roads, mutual interference between these radars will become more severe, increasing the probability of interference and the average power of the interference, leading to reduced detection accuracy. In severe cases, interference can create false targets within the radar's spectrum, ultimately leading to misjudgments.

[0003] Most existing automotive millimeter-wave radar interference removal methods require modern optimization algorithms to optimize the model, which is computationally intensive and difficult to implement in real time. Traditional filter-based methods, while not requiring optimization algorithms, suffer from poor performance and can leave residual interference signals. Summary of the Invention

[0004] In view of the problems existing in the prior art, the purpose of the present invention is to provide a method for eliminating automotive millimeter-wave radar interference signals based on sparse low-rank decomposition, so as to solve the problems of target signal loss and interference signal residue during the interference elimination process.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A method for eliminating interference signals from automotive millimeter-wave radars based on sparse low-rank decomposition is provided. The method obtains a radar signal containing an interference signal and processes the radar signal as follows:

[0007] Step 1: Initialize parameters: number of iterations K, relaxation parameter δ, standard deviation of interference signal σ i , the standard deviation σ of the radar signal x ;

[0008] Model the radar signal without interference as x and the radar signal with interference as y;

[0009] Step 2: Introduce the relaxation parameter δ to the radar signal y k In the above example, iterative regularization is performed to obtain the radar signal y k+1 :

[0010]

[0011] Among them, y k 、y k+1 are the radar signals after the k-th and k+1-th iterative processing of radar signal y, is the k-th estimated value of the radar signal x;

[0012] Step 3: Based on radar signal y k+1 Get an estimate of the standard deviation of the interference signal where γ is the scale factor; using this estimate Update the interference signal standard deviation σ i ;

[0013] Step 4: Transform the radar signal y k+1 Perform singular value decomposition:

[0014] (U k+1 ,Σ k+1 ,V k+1 )=SVD(y k+1 )

[0015] Among them, U k+1 and V k+1 is a unitary matrix, Σ k+1 is the singular value matrix;

[0016] Step 5: Based on the standard deviation σ of the interference signal i and the singular value matrix Σ k+1 Get an estimate of the standard deviation of the radar signal Use this estimate Update radar signal standard deviation σ x ;

[0017] Based on the standard deviation of the interference signal σ i and radar signal standard deviation σ x Calculate the constraint parameter τ k+1 :

[0018]

[0019] Step 6: Use the constraint parameter τ k+1 For the singular value matrix Σ k+1 Perform soft threshold processing to obtain the singular matrix estimate

[0020]

[0021] in, is the constraint parameter τ k+1Threshold function of

[0022] Step 7: Use the singular value matrix estimate to estimate the radar signal x and obtain the k+1th estimate of the radar signal

[0023] Step 8. Repeat steps 2-7 until the number of iterations reaches K times, completing the interference signal elimination work and obtaining the radar signal X after the interference signal is eliminated:

[0024]

[0025] When initializing parameters in step 1:

[0026] The standard deviation of the interference signal σ i , standard deviation σ of radar signal x Initialized to 0.

[0027] After adopting the above scheme, the singular value matrix of the radar signal is soft-thresholded, so that the radar target signal with large singular values ​​is retained, and the interference signal component with small singular values ​​is set to zero. The retained target signal singular values ​​are used to estimate the target signal. During the iterative regularization process, the partially recovered target signal The interference signal is substituted back into the radar signal to reduce the power ratio of the interference signal. During the iteration process, the power of the interference signal with low singular value is continuously reduced, and the target signal with high singular value is restored.

[0028] Therefore, the present invention considers the standard deviation of the interference signal and the radar signal when eliminating interference signals, eliminating target signal loss and ensuring a complete and accurate target signal. Furthermore, the present invention does not require modern algorithms to optimize the model, offering high performance and no prior knowledge. Therefore, it can be implemented on automotive millimeter-wave radars, demonstrating its potential for application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is the block diagram of the automotive millimeter wave radar system;

[0030] Figure 2 This is a schematic diagram of range-Doppler-azimuth three-dimensional radar data;

[0031] Figure 3 Schematic diagram of interference signal elimination according to the present invention;

[0032] Figure 4 This is the range Doppler spectrum before and after interference elimination in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0033] The automotive millimeter wave radar used in the present invention is a linear frequency modulation continuous wave radar, such as Figure 1 As shown in the figure, this automotive millimeter-wave radar includes a waveform generator, an oscillator, a transmitting antenna, a receiving antenna, a mixer, a low-pass filter, an analog-to-digital converter (A / D), and a baseband signal processor. The waveform generator generates a waveform signal, which is processed by the oscillator and then transmitted through the transmitting antenna. Upon encountering a target, the radar returns a target signal. The radar receives the target signal through the receiving antenna, mixes the received target signal with the transmitted waveform, and then performs low-pass filtering and analog-to-digital conversion to generate the radar signal.

[0034] like Figure 2 As shown in the figure, the radar signal (Chirp) is stored in the form of a matrix in the automotive millimeter-wave radar. When the automotive millimeter-wave radar is a multi-antenna system, the received data of multiple receiving array elements are stored in multiple channels (Channel). The sampling data of a single Chirp signal is stored in the form of columns (fast time dimension), and the periodic sampling data of multiple Chirp signals are stored in the form of rows (slow time dimension).

[0035] like Figure 3 As shown, the present invention provides a method for eliminating automotive millimeter-wave radar interference signals based on sparse low-rank decomposition. The radar data of a single array element (channel) in the radar signal is extracted as a single-processed radar data matrix (i.e., radar signal y below), and then the data matrix is ​​sent to the interference cancellation system for the following processing:

[0036] Step 1: Initialize the interference cancellation system parameters: number of iterations K, relaxation parameter δ, standard deviation of the interference signal σ i , the standard deviation σ of the radar signal x .

[0037] The number of iterations is selected based on the requirements of radar interference suppression. According to experience, about 15 iterations can achieve a good interference suppression effect. The relaxation parameter affects the convergence speed of interference elimination. Based on experience, the relaxation parameter can be selected to be around 0.65. When the power of the interference signal changes, the relaxation parameter can be appropriately increased. At initialization, the standard deviation σ of the interference signal i , standard deviation σ of radar signal x Initialized to 0.

[0038] Model the radar signal without interference as x and the radar signal with interference as y.

[0039] Step 2: Introduce the relaxation parameter δ to the radar signal y k In the above example, iterative regularization is performed to obtain the radar signal y k+1 :

[0040]

[0041] Among them, yk 、y k+1 are the radar signals after the k-th and k+1-th iterative processing of radar signal y, is the k-th estimated value of the radar signal x;

[0042] Step 3: Based on radar signal y k+1 Get an estimate of the standard deviation of the interference signal where γ is the scale factor; using this estimate Update the interference signal standard deviation σ i ;

[0043] Step 4: Transform the radar signal y k+1 Perform singular value decomposition:

[0044] (U k+1 ,Σ k+1 ,V k+1 )=SVD(y k+1 )

[0045] Among them, U k+1 and V k+1 is a unitary matrix, Σ k+1 is the singular value matrix;

[0046] Step 5: Based on the standard deviation of the interference signal σ i and the singular value matrix Σ k+1 Get an estimate of the standard deviation of the radar signal Use this estimate Update radar signal standard deviation σ x ;

[0047] Based on the standard deviation of the interference signal σ i and radar signal standard deviation σ x Calculate the constraint parameter τ k+1 :

[0048]

[0049] Step 6: Use the constraint parameter τ k+1 For the singular value matrix Σ k+1 Perform soft threshold processing to obtain the singular matrix estimate

[0050]

[0051] in, is the constraint parameter τ k+1 Threshold function of

[0052] Step 7: Use the singular value matrix estimate to estimate the radar signal x and obtain the k+1th estimate of the radar signal

[0053] Step 8. Repeat steps 2-7 until the number of iterations reaches K times, completing the interference signal elimination work and obtaining the radar signal X after the interference signal is eliminated:

[0054]

[0055] The present invention performs soft threshold processing on the singular value matrix of the radar signal, so that the radar target signal with large singular values ​​is retained and the interference signal component with small singular values ​​is set to zero. The retained target signal singular values ​​are used to estimate the target signal. During the iterative regularization process, the partially recovered target signal The signal is then substituted back into the interfering radar signal to reduce the power of the interfering signal. During the iteration process, the power of the interfering signal with low singular values ​​is continuously reduced, and the target signal with high singular values ​​is recovered.

[0056] In order to verify the effectiveness of the automotive millimeter-wave radar interference elimination method based on sparse low-quality decomposition provided by the present invention, this embodiment uses two automotive millimeter-wave radars operating at a center frequency of 77 GHz to collect data under radar interference conditions as an example.

[0057] Two automotive millimeter-wave radars use different frequency modulation parameters. A moving car is placed in front of the primary radar (the first radar) as a target of interest. An interference source (the second radar) is placed in front of the primary radar to generate a strong interference signal. This example demonstrates a general interference experiment setup.

[0058] like Figure 4 As shown in (a), after the automobile millimeter-wave radar generates interference, the power of the interference signal in the range Doppler spectrum increases significantly, reaching 38 dBm. At this time, the target signal-to-interference ratio decreases and cannot be detected.

[0059] The method provided by the present invention is used to eliminate interference in radar signals, such as Figure 4 (b), the interference power drops to -15dBm, and the target is clearly visible.

[0060] In summary, the present invention considers the standard deviation of the interference signal and the radar signal when eliminating interference signals, thus preventing target signal loss and ensuring a complete and accurate target signal. Furthermore, the present invention does not require modern algorithms to optimize the model, offering excellent performance and the absence of prior knowledge. Therefore, it can be implemented on automotive millimeter-wave radars, offering enhanced application value.

[0061] The above description is merely an embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any minor modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for eliminating automotive millimeter-wave radar interference signals based on sparse low-rank decomposition, characterized by: The method acquires a radar signal containing an interference signal and processes the radar signal as follows: Step 1: Initialize parameters: number of iterations K, relaxation parameter δ, standard deviation of interference signal σ i , the standard deviation σ of the radar signal x ; Model the radar signal without interference as x and the radar signal with interference as y; Step 2: Introduce the relaxation parameter δ to the radar signal y k In the above example, iterative regularization is performed to obtain the radar signal y k+1 : Among them, y k 、y k+1 are the radar signals after the k-th and k+1-th iterative processing of radar signal y, is the k-th estimated value of the radar signal x; Step 3: Based on radar signal y k+1 Get an estimate of the standard deviation of the interference signal where γ is the scale factor; using this estimate Update the interference signal standard deviation σ i ; Step 4: Transform the radar signal y k+1 Perform singular value decomposition: (U k+1 ,Σ k+1 ,V k+1 )=SVD(y k+1 ) Among them, U k+1 and V k+1 is a unitary matrix, Σ k+1 is the singular value matrix; Step 5: Based on the standard deviation σ of the interference signal i and the singular value matrix Σ k+1 Get an estimate of the standard deviation of the radar signal Use this estimate Update radar signal standard deviation σ x ; Based on the standard deviation of the interference signal σ i and radar signal standard deviation σ x Calculate the constraint parameter τ k+1 : Step 6: Use the constraint parameter τ k+1 For the singular value matrix Σ k+1 Perform soft threshold processing to obtain the singular matrix estimate in, is the constraint parameter τ k+1 Threshold function of Step 7: Use the singular value matrix estimate to estimate the radar signal x and obtain the k+1th estimate of the radar signal Step 8. Repeat steps 2-7 until the number of iterations reaches K times, completing the interference signal elimination work and obtaining the radar signal X after the interference signal is eliminated:

2. The method for eliminating automobile millimeter-wave radar interference signals based on sparse low-rank decomposition according to claim 1, characterized in that: When initializing parameters in step 1: The standard deviation of the interference signal σ i , standard deviation σ of radar signal x Initialized to 0.

Images