A millimeter wave radar based super-resolution angle estimation method

Abstract

The application discloses a millimeter wave radar-based super-resolution angle estimation method, which comprises the following steps: obtaining multi-element sampling data by receiving echo signals of a transmitting antenna by a receiving antenna; adding a Hanning window in a distance dimension and a speed dimension respectively; performing multi-channel non-coherent accumulation to improve signal-to-noise ratio; performing two-dimensional CFAR detection on the two-dimensional matrix data after non-coherent accumulation, performing CA-CFAR detection in the distance dimension and OS-CFAR detection in the speed dimension respectively, and obtaining matrix data containing a target; resolving and aggregating target points in a point trace condensation mode, so as to form a unique target point trace, and saving distance unit information and speed unit information of the target; and estimating a target azimuth by using an improved forward-backward Root-MUSIC algorithm on the multi-element data after compensation. The method considers that the forward-backward direction only uses the covariance matrix on the main diagonal line, and the application has a significant improvement in angle resolution and angle measurement accuracy.

Description

Technical Field

[0001] This invention belongs to the field of radar technology and relates to a super-resolution angle estimation method based on millimeter-wave radar. Background Technology

[0002] With the rapid development of millimeter-wave devices, high-speed digital signal processors, and FMCW (Frequency Modulated Continuous Wave) radar signal processing technology, millimeter-wave radar multi-target detection technology is now widely used in autonomous driving systems. Autonomous driving systems primarily detect road targets, including stationary targets, moving targets, and multiple moving targets. Compared to ultrasonic radar, lidar, and automotive optical sensors (such as cameras), millimeter-wave radar, with its superior characteristics such as long detection range, high resolution, less susceptibility to complex weather conditions like fog, rain, and snow, and all-weather, all-time operation, is more suitable for autonomous driving systems and is gradually becoming a key sensor for ADAS (Advanced Driver Assistance Systems) such as adaptive cruise control, automatic emergency braking, and forward collision avoidance. In conclusion, research on multi-target detection using millimeter-wave radar is essential for obtaining more accurate target information.

[0003] Multiple-input multiple-output (MIMO) radar is a novel radar concept. Based on spatial spectrum estimation, MIMO radar can determine the arrival directions of multiple spatial signals within a certain spatial region, thereby estimating the target's azimuth. While angle estimation algorithms are computationally intensive, their high-resolution characteristics have made them a key research area in radar technology.

[0004] With the continuous advancement of millimeter-wave radar in high-resolution and high-precision detection, MIMO radar systems and DOA super-resolution algorithms have attracted increasing attention from scholars and experts. Early DOA estimation algorithms such as MUSIC and ESPRIT, being based on subspace, suffer from limitations when the incident signal is correlated. The rank of the eigenvalue decomposition of the data covariance matrix is ​​less than the number of signal sources, thus these algorithms cannot accurately distinguish coherent signals. To address this issue, Gunjal proposed using a new matrix to eliminate or reduce the correlation between signal sources, then obtaining accurate angle estimates through spectral peak estimation. Yao Xintong proposed an improved MUSIC algorithm based on subspace projection, combining the improved algorithm with a spatial smoothing algorithm. This algorithm maintains high resolution and good direction-finding performance even with low signal-to-noise ratios and small snapshot numbers. Summary of the Invention

[0005] To address the problem that the MUSIC algorithm cannot accurately distinguish coherent signals when faced with coherent sources, this invention provides the following technical solution:

[0006] A super-resolution angle estimation method based on millimeter-wave radar includes the following steps:

[0007] Step 1: Set the parameters for transmitting sawtooth wave signals using the transmitting antenna, and use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data;

[0008] Step 2: Rearrange the multi-element sampling data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data respectively. Add Hanning windows to the distance dimension and velocity dimension respectively.

[0009] Step 3: Based on the sampling points and pulse count, perform a two-dimensional FFT transformation on the three-dimensional distance and velocity data added to the Hanning window to obtain two-dimensional matrix data, and perform multi-channel incoherent accumulation to improve the signal-to-noise ratio;

[0010] Step 4: Perform two-dimensional CFAR detection on the two-dimensional matrix data after incoherent accumulation. Perform CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target.

[0011] Step 5: Since the target's matrix data still exists in the form of a point cloud, the target points are distinguished and aggregated using the point trace aggregation method to form a unique target point trace, and the distance cell information and velocity cell information of the target are saved.

[0012] Step 6: Perform Doppler compensation on the velocity cell where the target is located, and use the improved forward and backward Root-MUSIC algorithm to estimate the azimuth angle of the target on the compensated multi-array data.

[0013] Furthermore: both the transmitting antenna and the receiving antenna are equidistant linear arrays;

[0014] The equivalent antenna array has M elements. Assuming there are K targets in different orientations, the received signal is represented as:

[0015] X(t) = [X1(t), X2(t), ... X i (t),...X M (t)] T ,i=1,2,...M (1)

[0016] Where, x i (t)=A×S(t)+n(t),A=[α1(θ1),α2(θ2),...α K (θ K S is the M×K dimensional receiving array steering vector matrix, where S = [S1(t), S2(t), ... S...]. K[n(t)] is a K×1 dimensional vector of K received signals in complex envelope form, N=[n1(t),n2(t),...n k [(t)] is the M×1 dimensional noise vector in the array signal.

[0017] Furthermore, the specific steps for rearranging the multi-element sampling data to obtain three-dimensional matrix data are as follows: the sampling data is rearranged according to the number of elements, the number of sampling points, and the number of pulses to obtain three-dimensional matrix data.

[0018] Furthermore: the specific process of the two-dimensional CFAR detection is as follows:

[0019] Set the number of reference elements, the number of guard elements, and the scale factor. Use the CA-CFAR algorithm on the distance dimension and the OS-CFAR algorithm on the velocity dimension to determine whether the element to be detected is a target point.

[0020] The CA-CFAR algorithm specifically involves: averaging the distance-dimensional FFT data of the target cell; calculating the background noise power; multiplying the noise power by a scale factor to obtain a detection threshold estimate; and comparing the detection threshold estimate with the data of the target cell.

[0021] If the data of the unit to be detected is greater than or equal to the estimated detection threshold, it indicates that a target exists; if the data of the unit to be detected is less than the estimated detection threshold, it indicates that a target exists; otherwise, it indicates that there is no target.

[0022] The OS-CFAR algorithm is as follows: After sorting the velocity dimension FFT data of the unit to be detected, the kth value is taken as the background noise power value. The background noise power value is multiplied by the scale factor to obtain the detection threshold estimate and compared with the data of the unit to be detected. If the data of the unit to be detected is greater than or equal to the detection threshold estimate, it means that there is a target. If the data of the unit to be detected is less than the detection threshold estimate, it means that there is no target.

[0023] The final output is a two-dimensional matrix of data containing the target.

[0024] Furthermore: the process of performing Doppler compensation on the velocity cell where the target is located, and then using the improved forward-backward Root-MUSIC method to estimate the azimuth angle of the target using the compensated multi-array data, is as follows:

[0025] Step 601: Due to the Doppler effect of the moving target, a Doppler phase shift will be generated. This Doppler shift will cause inaccurate angle measurement. Extract the velocity information of the target from the two-dimensional matrix, calculate its phase shift and compensate it back into the two-dimensional matrix.

[0026]

[0027] Where n represents the index of the velocity unit where the current velocity is located, and N represents all velocity units;

[0028] Step 602: Based on the target's velocity element information, extract the data of the velocity dimension of the target in the three-dimensional matrix. Rearrange the data into a two-dimensional matrix according to the number of elements and sampling points. Divide the two-dimensional matrix into p overlapping forward subarrays and p overlapping backward subarrays. Each subarray has m elements, thus obtaining the forward data X. f (k) and backward data X b (k). Cross-correlation is performed on the autocorrelation matrices of all subarray data, and then summation and parity are applied to obtain the improved forward and backward spatial smoothing matrix, which is expressed as:

[0029]

[0030] Step 603: Let J be an m×m reverse identity matrix, such that... in Indicates R fb The complex conjugate of J, the J matrix is ​​as follows:

[0031]

[0032] Step 604: Perform eigenvalue decomposition on the covariance matrix R to obtain multiple target eigenvalues ​​λ1, λ2, ... λ m And the corresponding feature vectors v1, v2, ... v m ;

[0033] Step 605: Take the top K largest eigenvalue vectors as the target eigenvectors, and the remaining mK eigenvectors as noise eigenvectors. Sort the noise eigenvectors according to their eigenvalues ​​to construct a noise subspace.

[0034] Step 606: The ratio between the antenna spacing and the radar wavelength is d. Let the angle of the target signal relative to the radar be θ, then the steering vector is: A = [1, e] -j2πdsin(θ) ,...e -j2π(m-1)dsin(θ) The spatial spectral function is constructed as follows:

[0035]

[0036] Step 607: Expand the denominator of the spatial spectral function to obtain:

[0037]

[0038] Setting the above equation to 0, there are 2(M-1) roots that appear in pairs. Among these roots, there are exactly K pairs of roots on the unit circle. In reality, there may be some errors. Therefore, by finding the K pairs of roots that are closest to the unit circle, we can obtain the estimated value of the target azimuth.

[0039] A super-resolution angle estimation device based on millimeter-wave radar includes the following steps:

[0040] Acquisition module: used to set the parameters for the transmitting antenna to transmit sawtooth wave signals, and to use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data;

[0041] The rearrangement module is used to rearrange the multi-element sampled data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data, respectively, and Hanning windows are added to the distance dimension and velocity dimension.

[0042] Transformation module: Based on the sampling points and pulse count, the three-dimensional distance and velocity data with Hanning window are transformed into two-dimensional matrix data by two-dimensional FFT, and multi-channel incoherent accumulation is performed to improve the signal-to-noise ratio;

[0043] Detection module: Used to perform two-dimensional CFAR detection on two-dimensional matrix data that has undergone incoherent accumulation, performing CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target;

[0044] Formation module: Used to distinguish and aggregate target points using point aggregation method, thereby forming a unique target point trace, and saving the distance unit information and velocity unit information of the target;

[0045] Compensation module: Used to perform Doppler compensation on the velocity cell where the target is located, and to estimate the azimuth angle of the target using an improved forward and backward Root-MUSIC algorithm on the compensated multi-array data.

[0046] This invention provides a super-resolution angle estimation method based on millimeter-wave radar. Based on a 77GHz millimeter-wave radar and combined with practical application scenarios, this method proposes an improved forward and backward Root-MUSIC algorithm on the basis of the existing MUSIC algorithm. This method not only solves the problem of the MUSIC algorithm's performance deteriorating sharply when facing coherent sources, but also effectively improves angular resolution. This method considers that the forward and backward directions only use the covariance matrix on the main diagonal. To fully utilize the receiving array, the covariance matrix is ​​improved and converted into a Hermite Toeplize matrix, making the new covariance matrix an unbiased estimate of the original covariance matrix. Performance comparison with traditional algorithms shows that this invention significantly improves both angle resolution and angle measurement accuracy. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating the steps of the method.

[0049] Figure 2 The result image of adding a two-dimensional FFT using the Hanning window;

[0050] Figure 3 The result of a 2D FFT without the Hanning window is shown.

[0051] Figure 4 This is a graph showing the CFAR detection results;

[0052] Figure 5 The result of dot aggregation;

[0053] Figure 6 The graph shows the estimation results of the three methods when the target is located at (-10°, 0°, 10°).

[0054] Figure 7 The graph shows the estimation results of the two methods when the target is located at (-5°, 0°, 5°).

[0055] Figure 8 The graph shows the transformation of the root mean square error for the two methods under different signal-to-noise ratios.

[0056] Figure 9 The graph shows the transformation of the root mean square error for the two methods under different sampling points.

[0057] Figure 10 This is the root distribution map of the target located at (0°, 5°) in the actual test scenario. Detailed Implementation

[0058] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0061] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0062] In the description of this invention, it should be understood that the orientation or positional relationship indicated by directional terms such as "front, back, up, down, left, right", "horizontal, vertical, horizontal" and "top, bottom" is generally based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. Unless otherwise stated, these directional terms do not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the scope of protection of this invention. The directional terms "inner" and "outer" refer to the inner and outer contours relative to the outline of each component itself.

[0063] For ease of description, spatial relative terms such as "above," "over," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation besides the orientation of the device as described in the figures. For example, if the device in the figures is inverted, a device described as "above" or "above" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0064] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.

[0065] Figure 1 Here is a diagram illustrating the steps of this method:

[0066] A super-resolution angle estimation method based on millimeter-wave radar includes the following steps:

[0067] Step 1: Set the parameters for transmitting sawtooth wave signals using the transmitting antenna, and use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data;

[0068] Step 2: Rearrange the multi-element sampling data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data respectively. Add Hanning windows to the distance dimension and velocity dimension respectively.

[0069] Step 3: Based on the sampling points and pulse count, perform a two-dimensional FFT transformation on the three-dimensional distance and velocity data with the Hanning window to obtain two-dimensional matrix data. Then, perform multi-channel incoherent accumulation to improve the signal-to-noise ratio. The windowed two-dimensional FFT signal is as follows: Figure 2 Unwindowed 2D FFT signal, such as Figure 3 ;

[0070] Step 4: Perform two-dimensional CFAR detection on the two-dimensional matrix data after incoherent accumulation. Perform CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target.

[0071] Step 5: Since the target's matrix data still exists in the form of a point cloud, the target points are distinguished and aggregated using the point trace aggregation method to form a unique target point trace, and the distance cell information and velocity cell information of the target are saved.

[0072] Step 6: Perform Doppler compensation on the velocity cell where the target is located, and use the improved forward and backward Root-MUSIC method to estimate the azimuth angle of the target on the compensated multi-array data.

[0073] The steps S1 / S2 / S3 / S4 / S5 / S6 are executed sequentially;

[0074] Furthermore: both the transmitting antenna and the receiving antenna are equidistant linear arrays; and the transmitting antenna includes M1 array elements, and the receiving antenna includes M2 array elements, with an equivalent number of array elements of M = M1M2.

[0075] The equivalent antenna array has M elements. Assuming there are K targets in different orientations, the received signal is represented as:

[0076] X(t) = [X1(t), X2(t), ... X i (t),...X M (t)] T ,i=1,2,...M (1)

[0077] Where, x i (t)=A×S(t)+n(t),A=[α1(θ1),α2(θ2),...α K (θ K S is the M×K dimensional receiving array steering vector matrix, where S = [S1(t), S2(t), ... S...]. K [n(t)] is a K×1 dimensional vector of K received signals in complex envelope form, N=[n1(t),n2(t),...n k [(t)] is the Mm1-dimensional noise vector in the array signal.

[0078] Furthermore, the process of rearranging the multi-element sampled data to obtain three-dimensional matrix data is as follows: The sampled data is rearranged according to the number of elements, sampling points, and pulses to obtain the three-dimensional matrix data. The effective difference frequency signal is obtained by multiplying the distance and velocity dimensions by a Hanning window, respectively.

[0079] Furthermore: the specific process of the two-dimensional CFAR detection is as follows:

[0080] Step 401: Set the number of reference cells, the number of guard cells, and the scale factor. Use the CA-CFAR algorithm on the distance dimension and the OS-CFAR algorithm on the velocity dimension to determine whether the cell to be detected is a target point.

[0081] Step 402: The CA-CFAR algorithm specifically involves: After averaging the distance-dimensional FFT data of the unit to be detected, calculating the background noise power; multiplying the noise power by the scale factor to obtain the detection threshold estimate; and comparing the detection threshold estimate with the data of the unit to be detected.

[0082] If the data of the unit to be detected is greater than or equal to the estimated detection threshold, it indicates that a target exists; if the data of the unit to be detected is less than the estimated detection threshold, it indicates that a target exists; otherwise, it indicates that there is no target.

[0083]

[0084] Where d is the power of the unit to be detected, α is the scale factor, and T CA This is the estimated background noise power value.

[0085] If the value of the unit to be detected is greater than the detection threshold, then H1 is assumed to be true, meaning the unit has a target; otherwise, the unit has no target.

[0086] The OS-CFAR algorithm is as follows: After sorting the velocity dimension FFT data of the unit to be detected, the kth value is taken as the background noise power value. The background noise power value is multiplied by the scale factor to obtain the detection threshold estimate and compared with the data of the unit to be detected. If the data of the unit to be detected is greater than or equal to the detection threshold estimate, it means that there is a target. If the data of the unit to be detected is less than the detection threshold estimate, it means that there is no target.

[0087] The final output is a two-dimensional matrix of data containing the target.

[0088]

[0089] Where d is the power of the unit to be detected, α is the scale factor, and T OS The k-th value is selected as the estimated background noise power value.

[0090] The result obtained after CFAR testing is shown in the figure below. Figure 4 ;

[0091] Furthermore, the process of using dot clustering to distinguish and aggregate target points to form a unique target dot pattern is as follows:

[0092] Step 501: For the signal after CFAR detection and processing, firstly merge and classify it in the range dimension. When the target is larger than the radar resolution, the target will split, generating multiple range cell values ​​in the range dimension. Determine whether the original target points are continuous in a certain range cell segment, and merge the continuous points together to complete the merging. If there are two batches of adjacent points, classify them according to the amplitude difference of the adjacent point data. If the difference is less than the threshold value, then it is the same target. If it is greater than the threshold value, it is two batches of targets.

[0093] Step 502: Next, merge and classify along the velocity dimension, following the same steps as along the distance dimension;

[0094] Step 503: After all points have been marked, output the point with the largest amplitude among the points as the target point. The effect after point aggregation is shown in the figure below. Figure 5 ;

[0095] Furthermore: the process of performing Doppler compensation on the velocity cell where the target is located, and then using the improved forward-backward Root-MUSIC method to estimate the azimuth angle of the target using the compensated multi-array data, is as follows:

[0096] Step 601: Due to the Doppler effect of the moving target, a Doppler phase shift will be generated. This Doppler shift will cause inaccurate angle measurement. Extract the velocity information of the target from the two-dimensional matrix, calculate its phase shift and compensate it back into the two-dimensional matrix.

[0097]

[0098] Where n represents the index of the velocity unit where the current velocity is located, and N represents all velocity units;

[0099] Step 602: Based on the target's velocity element information, extract the data of the velocity dimension of the target in the three-dimensional matrix. Rearrange the data into a two-dimensional matrix according to the number of elements and sampling points. Divide the two-dimensional matrix into p overlapping forward subarrays and p overlapping backward subarrays. Each subarray has m elements, thus obtaining the forward data X. f (k) and backward data X b (k). Cross-correlation is performed on the autocorrelation matrices of all subarray data, and then summation and parity are applied to obtain the improved forward and backward spatial smoothing matrix, which is expressed as:

[0100]

[0101] The formula can be rewritten as:

[0102]

[0103] make

[0104] We can obtain that if m > K, rank{R} s If}=1, then when p≥K, rank{R} s f Since} = K, the K coherent signal sources can be treated as independent signal sources for the subsequent Root-MUSIC processing.

[0105] Step 603: Let J be an m×m reverse identity matrix, such that... in Indicates R fb The complex conjugate of J, the J matrix is ​​as follows:

[0106]

[0107] The purpose of this is to make R a Hermite Toeplize matrix, so that R is R fb Unbiased estimation can further improve its detection performance.

[0108] Step 604: Perform eigenvalue decomposition on the covariance matrix R to obtain multiple target eigenvalues ​​λ1, λ2, ... λ m And the corresponding feature vectors v1, v2, ... v m ;

[0109] Step 605: Take the top K largest eigenvalue vectors as the target eigenvectors, and the remaining mK eigenvectors as noise eigenvectors. Sort the noise eigenvectors according to their eigenvalues ​​to construct a noise subspace.

[0110] Step 606: The ratio between the antenna spacing and the radar wavelength is d. Let the angle of the target signal relative to the radar be θ, then the steering vector is: A = [1, e] -j2πdsin(θ) ,...e -j2π(m-1)dsin(θ) The spatial spectral function is constructed as follows:

[0111]

[0112] Step 607: Expand the denominator of the spatial spectral function to obtain:

[0113]

[0114] Setting the above equation to 0, there are 2(M-1) roots that appear in pairs. Among these roots, there are exactly K pairs of roots on the unit circle. In reality, there may be some errors. Therefore, by finding the K pairs of roots that are closest to the unit circle, we can obtain the estimated value of the target azimuth.

[0115] A super-resolution angle estimation device based on millimeter-wave radar includes the following steps:

[0116] Acquisition module: used to set the parameters for the transmitting antenna to transmit sawtooth wave signals, and to use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data;

[0117] The rearrangement module is used to rearrange the multi-element sampled data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data, respectively, and Hanning windows are added to the distance dimension and velocity dimension.

[0118] Transformation module: Used to perform a two-dimensional FFT transformation on the three-dimensional distance and velocity data added to the Hanning window based on the sampling points and pulse number to obtain two-dimensional matrix data, and perform multi-channel incoherent accumulation to improve the signal-to-noise ratio;

[0119] Detection module: Used to perform two-dimensional CFAR detection on two-dimensional matrix data that has undergone incoherent accumulation, performing CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target;

[0120] Formation module: Used to distinguish and aggregate target points using point aggregation method, thereby forming a unique target point trace, and saving the distance unit information and velocity unit information of the target;

[0121] Compensation module: Used to perform Doppler compensation on the velocity cell where the target is located, and to estimate the azimuth of the target using the improved forward and backward Root-MUSIC method on the compensated multi-array data.

[0122] To further illustrate the effects of the present invention, comparative experiments can be conducted:

[0123] (1) Simulation conditions:

[0124] Assume the number of transmitting antenna elements M1 = 2, the number of receiving antenna elements M2 = 4, the element spacing d = λ / 2, the starting frequency of the transmitted signal is 77 GHz, the number of sampling points is 256, the sampling rate is 8 MHz, the number of pulses is 128, the pulse repetition period is 32 μs, the slope is 48 MHz / μs, and the signal-to-noise ratio is 10 dB.

[0125] Simulation 1: Three coherent signals are set at azimuth angles of (-10°, 0°, 10°) on the radar. Azimuth angle estimation is performed using the MUSIC algorithm, the forward / backward spatial smoothing MUSIC algorithm, and the improved forward / backward root-MUSIC algorithm. The search range of the spatial spectrum is -60° to 60°, and the search step size is 0.1°. The results are as follows. Figure 6As shown in the figure, both the forward and backward spatial smoothing MUSIC algorithm and the improved forward and backward root-MUSIC algorithm can distinguish the target's angular information relatively well, but the MUSIC algorithm has completely failed.

[0126] Simulation 2: Three coherent signals are set at azimuth angles of (-5°, 0°, 5°) on the radar. Azimuth angle estimation is performed using the forward and backward spatial smoothing MUSIC algorithm and the improved forward and backward root-MUSIC algorithm. The search range of the spatial spectrum is -60° to 60°, and the search step size is 0.1°. The results are as follows. Figure 7 As shown in the figure, the forward and backward spatial smoothing MUSIC algorithm is no longer able to resolve three targets, with only two peak points appearing. In contrast, the improved forward and backward root-MUSIC algorithm can better distinguish the angular information of the targets.

[0127] Simulation 3: Two coherent signals are set, located at azimuth angles of (0°, 10°) on the radar. Azimuth angle estimation is performed using both the forward and backward spatial smoothing MUSIC algorithm and an improved forward and backward root-MUSIC algorithm. The root mean square error (RSEM) of the two decorrelation super-resolution angle estimation algorithms under different signal-to-noise ratios is shown below. Figure 8 As shown in the figure, as the SNR increases, the RMSE values ​​of the two decorrelation algorithms gradually decrease until they are very close. However, the improved forward and backward Root-MUSIC algorithm proposed in this invention consistently performs worse than the forward and backward spatial smoothing MUSIC algorithm.

[0128] Simulation 4: Two coherent signals are set up, located at azimuth angles of (0°, 10°) on the radar. Azimuth angle estimation is performed using the forward and backward spatial smoothing MUSIC algorithm and an improved forward and backward root-MUSIC algorithm. The root mean square error (RSEM) of the two decorrelation super-resolution angle estimation algorithms under different sampling point conditions is as follows: Figure 9 As shown, with the increase of sampling points, the RMSE values ​​of the two decoherence algorithms gradually decrease and their accuracy gradually increases. It can be seen that the algorithm proposed in this invention has a lower estimation error than the forward and backward spatial smoothing MUSIC algorithm.

[0129] Actual data processing and analysis:

[0130] Using the three angle measurement methods described above, the measured data of the 77GHz millimeter-wave radar in an indoor environment were processed. The processing results for the angles reversed at 0° and 5° positions are as follows: Figure 10 As shown in the diagram, the root distribution plot reveals two pairs of roots closest to the unit circle. The angles measured using the Root-MUSIC method before and after the improvement are 0.8° and 5.2°.

[0131] Simulation and experimental results show that the method proposed in this invention has better angle resolution and angle measurement accuracy.

[0132] This invention provides a super-resolution angle estimation method based on millimeter-wave radar. There are many ways to implement this technical solution. The above description is only the best embodiment of this invention. However, the scope of protection of this invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of this invention should be included within the scope of protection of this invention.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A super-resolution angle estimation method based on millimeter-wave radar, characterized in that: Includes the following steps: Step 1: Set the parameters for transmitting sawtooth wave signals using the transmitting antenna, and use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data; Step 2: Rearrange the multi-element sampling data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data respectively. Add Hanning windows to the distance dimension and velocity dimension respectively. Step 3: Based on the sampling points and pulse count, perform a two-dimensional FFT transformation on the three-dimensional distance and velocity data added to the Hanning window to obtain two-dimensional matrix data, and perform multi-channel incoherent accumulation to improve the signal-to-noise ratio; Step 4: Perform two-dimensional CFAR detection on the two-dimensional matrix data after incoherent accumulation. Perform CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target. Step 5: Since the target's matrix data still exists in the form of a point cloud, the target points are distinguished and aggregated using the point trace aggregation method to form a unique target point trace, and the distance cell information and velocity cell information of the target are saved. Step 6: Perform Doppler compensation on the velocity cell containing the target, and use the improved forward-backward Root-MUSIC algorithm to estimate the azimuth angle of the target using the compensated multi-array data. The specific process is as follows: Step 601: Due to the Doppler effect of the moving target, a Doppler phase shift will be generated. This Doppler shift will cause inaccurate angle measurement. Extract the velocity information of the target from the two-dimensional matrix, calculate its phase shift and compensate it back into the two-dimensional matrix. （4） Where n represents the index of the velocity unit where the current velocity is located, and N represents all velocity units; Step 602: Based on the target's velocity element information, extract the data of the velocity dimension of the target in the three-dimensional matrix. Rearrange the data into a two-dimensional matrix according to the number of elements and sampling points. Divide the two-dimensional matrix into p overlapping forward subarrays and p overlapping backward subarrays, with each subarray containing m elements, to obtain the forward data. and backward data The autocorrelation matrices of all subarray data are cross-correlated, and then summed and flattened to obtain the improved forward and backward spatial smoothing matrix, which is expressed as: （5） Step 603: Let for The reverse identity matrix, making ,in Indicates to The complex conjugate, The matrix is ​​as follows: （7） Step 604: For the covariance matrix Perform eigenvalue decomposition to obtain multiple target eigenvalues. and the corresponding feature vectors ; Step 605: Take the top K largest eigenvalue vectors as the target eigenvectors, and the remaining mK eigenvectors as noise eigenvectors. Sort the noise eigenvectors according to their eigenvalues ​​to construct a noise subspace. ; Step 606: The ratio between the antenna spacing and the radar wavelength is d. Let the angle of the target signal relative to the radar be... Then the steering vector is: The spatial spectral function is constructed as follows: （8） Step 607: Expand the denominator of the spatial spectral function to obtain: （9） Setting the above equation to 0, there are 2(M-1) roots that appear in pairs. Among these roots, there are exactly K pairs of roots on the unit circle. In reality, there may be some errors. Therefore, by finding the K pairs of roots that are closest to the unit circle, we can obtain the estimated value of the target azimuth.

2. The super-resolution angle estimation method based on millimeter-wave radar according to claim 1, characterized in that: Both the transmitting antenna and the receiving antenna are equidistant linear arrays. The equivalent antenna array has M elements. Assuming there are K targets in different orientations, the received signal is represented as: （1） in, , for The dimensional receiver array steering vector matrix, For the received signal in the form of K complex envelopes 3D vector For array signals 3D noise vector.

3. The super-resolution angle estimation method based on millimeter-wave radar according to claim 1, characterized in that: The specific steps for rearranging multi-element sampling data to obtain three-dimensional matrix data are as follows: the sampling data is rearranged according to the number of array elements, the number of sampling points, and the number of pulses to obtain three-dimensional matrix data.

4. The super-resolution angle estimation method based on millimeter-wave radar according to claim 1, characterized in that: The specific process of the two-dimensional CFAR detection is as follows: Set the number of reference elements, the number of guard elements, and the scale factor. Use the CA-CFAR algorithm on the distance dimension and the OS-CFAR algorithm on the velocity dimension to determine whether the element to be detected is a target point. The CA-CFAR algorithm specifically involves: averaging the distance-dimensional FFT data of the target cell; calculating the background noise power; multiplying the noise power by a scale factor to obtain a detection threshold estimate; and comparing the detection threshold estimate with the data of the target cell. If the data of the unit to be detected is greater than or equal to the estimated detection threshold, it indicates that a target exists; if the data of the unit to be detected is less than the estimated detection threshold, it indicates that a target exists; otherwise, it indicates that there is no target. The OS-CFAR algorithm is as follows: After sorting the velocity dimension FFT data of the unit to be detected, the kth value is taken as the background noise power value. The background noise power value is multiplied by the scale factor to obtain the detection threshold estimate and compared with the data of the unit to be detected. If the data of the unit to be detected is greater than or equal to the detection threshold estimate, it means that there is a target. If the data of the unit to be detected is less than the detection threshold estimate, it means that there is no target. The final output is a two-dimensional matrix of data containing the target.

5. A super-resolution angle estimation device based on millimeter-wave radar, characterized in that: Includes the following steps: Acquisition module: used to set the parameters for the transmitting antenna to transmit sawtooth wave signals, and to use the receiving antenna to receive the echo signals from the transmitting antenna to obtain multi-element sampling data; The rearrangement module is used to rearrange the multi-element sampled data to obtain three-dimensional matrix data, which are distance dimension data, velocity dimension data and angle dimension data, respectively, and Hanning windows are added to the distance dimension and velocity dimension. Transformation module: Based on the sampling points and pulse count, the three-dimensional distance and velocity data with Hanning window are transformed into two-dimensional matrix data by two-dimensional FFT, and multi-channel incoherent accumulation is performed to improve the signal-to-noise ratio; Detection module: Used to perform two-dimensional CFAR detection on two-dimensional matrix data that has undergone incoherent accumulation, performing CA-CFAR detection in the distance dimension and OS-CFAR detection in the velocity dimension to obtain matrix data containing the target; Formation module: Used to distinguish and aggregate target points using point aggregation method, thereby forming a unique target point trace, and saving the distance unit information and velocity unit information of the target; Compensation module: used to perform Doppler compensation on the velocity cell where the target is located, and use the improved forward and backward Root-MUSIC algorithm to estimate the azimuth of the target on the compensated multi-array data. The process of performing Doppler compensation on the velocity cell where the target is located, and then using the improved forward-backward Root-MUSIC algorithm to estimate the azimuth angle of the target using the compensated multi-array data is as follows: Step 601: Due to the Doppler effect of the moving target, a Doppler phase shift will be generated. This Doppler shift will cause inaccurate angle measurement. Extract the velocity information of the target from the two-dimensional matrix, calculate its phase shift and compensate it back into the two-dimensional matrix. （4） Where n represents the index of the velocity unit where the current velocity is located, and N represents all velocity units; Step 602: Based on the target's velocity element information, extract the data of the velocity dimension of the target in the three-dimensional matrix. Rearrange the data into a two-dimensional matrix according to the number of elements and sampling points. Divide the two-dimensional matrix into p overlapping forward subarrays and p overlapping backward subarrays, with each subarray containing m elements, to obtain the forward data. and backward data The autocorrelation matrices of all subarray data are cross-correlated, and then summed and flattened to obtain the improved forward and backward spatial smoothing matrix, which is expressed as: （5） Step 603: Let for The reverse identity matrix, making ,in Indicates to The complex conjugate, The matrix is ​​as follows: （7） Step 604: For the covariance matrix Perform eigenvalue decomposition to obtain multiple target eigenvalues. and the corresponding feature vectors ; Step 605: Take the top K largest eigenvalue vectors as the target eigenvectors, and the remaining mK eigenvectors as noise eigenvectors. Sort the noise eigenvectors according to their eigenvalues ​​to construct a noise subspace. ; Step 606: The ratio between the antenna spacing and the radar wavelength is d. Let the angle of the target signal relative to the radar be... Then the steering vector is: The spatial spectral function is constructed as follows: （8） Step 607: Expand the denominator of the spatial spectral function to obtain: （9） Setting the above equation to 0, there are 2(M-1) roots that appear in pairs. Among these roots, there are exactly K pairs of roots on the unit circle. In reality, there may be some errors. Therefore, by finding the K pairs of roots that are closest to the unit circle, we can obtain the estimated value of the target azimuth.

Images