CN109975807B - Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar - Google Patents
Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar Download PDFInfo
- Publication number
- CN109975807B CN109975807B CN201910238433.3A CN201910238433A CN109975807B CN 109975807 B CN109975807 B CN 109975807B CN 201910238433 A CN201910238433 A CN 201910238433A CN 109975807 B CN109975807 B CN 109975807B
- Authority
- CN
- China
- Prior art keywords
- signal
- matrix
- subspace
- transmitting
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
- G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a dimension reduction subspace angle measurement method suitable for the field of millimeter wave vehicle-mounted radar target detection, which comprises the following steps: firstly, a beam domain MUSIC algorithm is adopted to reduce the calculation complexity and the memory occupation, and the prior information is utilized to optimize the beam forming matrix design; secondly, a new MUSIC estimator is proposed, reflecting more detailed scale information about the signal subspace; in addition, aiming at the condition that two adjacent signals are coherent under single snapshot, the mathematical expression of the covariance matrix of the received signal estimation sample is modified so as to enhance the orthogonality of a noise subspace and a signal subspace and keep the Toeplitz structure of the signal subspace, and the estimation precision of the covariance matrix of the sample is improved under the condition of full rank of each sub-beam space by adopting a sub-beam space averaging method.
Description
Technical Field
The invention belongs to the field of radar signal processing, and particularly relates to a dimension reduction subspace angle measurement method suitable for a millimeter wave vehicle-mounted radar.
Background
In recent years, the fields of automatic driving are vigorously developed, and automatic cruising, blind area detection, forward collision prevention and the like are typical application scenes. The millimeter wave radar system is an indispensable ring in the technical field of automatic driving, has good performance under severe weather environments such as fog, weak light and the like, and is not provided by other sensors such as laser radars, cameras and the like. A Frequency Modulated Continuous Wave (FMCW) radar, which is a millimeter wave radar system that obtains distance and speed information by frequency modulating a continuous wave, has been limited in its application to a small range for a long time in the past. In the nineties, the development of solid-state microwave millimeter wave devices and digital signal processing technology lays a foundation for the development of millimeter wave FMCW radars. The millimeter wave radar has extremely deep application value in the aspects of military use and civil use, and the advantages of the millimeter wave radar can be summarized as follows:
1. a large amount of bandwidth can be used, the distance measurement resolution ratio is improved, mutual interference is effectively eliminated, and no speed measurement blind area exists.
2. The wavelength is shorter, the beam width is narrow, the antenna gain is high, the spatial resolution can be improved, and meanwhile, the element size is small and the weight is light.
3. The atmospheric absorption effect is stronger than that of microwave, the attenuation is large, mutual interference is not easy to occur, and the electromagnetic pollution is reduced.
Based on the above advantages of FMCW, FMCW is widely used in vehicle radar systems.
In addition to the traditional detection of the target distance and speed, the vehicle-mounted millimeter wave radar also puts high requirements on the estimation of the target arrival angle. In the past decades, MUSIC algorithm has been considered as an effective subspace angle estimation algorithm due to its superior performance. However, a big problem of the MUSIC algorithm is that it usually includes eigenvalue decomposition and spectral peak search operations on array signal data received by a large number of sensors, and the amount of computation is very large, especially in a vehicle-mounted millimeter wave FMCW radar system, the number of DSP processing chips is limited, and very high requirements are imposed on the computation complexity and the memory occupation. In addition, the incoming wave direction estimation in the application scene of the vehicle-mounted radar is carried out at a specific distance where a target is located and a Doppler unit, and actually aims at received signal data of a single snapshot, when a plurality of adjacent targets exist, the received signals cannot be regarded as incoherent signals, the Toeplitz structure of the covariance matrix of the received signals is damaged, and a new requirement is also put forward on an angle measurement scheme.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problems, the invention provides a dimension reduction subspace angle measurement method suitable for a millimeter wave vehicle-mounted radar, which meets the requirements of low computation complexity and memory occupation on one hand, and has excellent angle measurement performance under the condition of single target or multiple targets on the other hand.
The technical scheme is as follows: in order to realize the purpose of the invention, the technical scheme adopted by the invention is as follows: to facilitate the description in this section, table 1 lists descriptions of all the parameters commonly used in the present invention.
Table 1 general parameter description
The invention discloses a dimensionality reduction subspace angle measurement method suitable for the field of millimeter wave vehicle-mounted radar target detection.
Step 1: establishing a space spectrum estimation mathematical model of the millimeter wave vehicle-mounted radar system to obtain mathematical expressions of a transmitting signal and a receiving signal;
step 2: expanding and establishing a three-dimensional data structure of the intermediate frequency signal after mixing the received signal under the condition of multiple receiving and transmitting on the basis of the step 1 by utilizing the number and the layout of transmitting antennas and receiving antennas required by a millimeter wave vehicle-mounted radar system;
and step 3: performing FFT on the three-dimensional received signal data obtained in the step 2 in a fast time dimension and a slow time dimension respectively to obtain the distance and speed information of the detected target vehicle and a received signal data vector Y of the antenna array on the corresponding distance Doppler unit;
and 4, step 4: calculating an optimized beam forming matrix B by using the azimuth angle range information of the vehicle to be detected by the vehicle radaroptConverting the received signal data vector Y of the antenna array on the specific range Doppler unit from the array element domain to the beam domain to obtain the received signal vector Y of the beam domainB;
And 5: using the beam domain received signal matrix Y obtained in step 4BComputing a sample covariance matrixAnd toCorrecting to obtain a corrected sample covariance matrix
Step 6: for the sample covariance matrix corrected in step 5Performing eigenvalue decomposition to obtain noise subspaceBy usingAndestablishing a new MUSIC estimator feMUSICPerforming spectral peak search to obtain an incoming wave direction estimate
Further, in step 1, the waveform of the transmission signal of the millimeter wave vehicle-mounted radar system is a group of carrier frequencies f0A linear frequency modulation continuous wave signal (LFMCW) with a certain sweep frequency bandwidth in a transmission period, a plurality of transmitting antennas transmit in time division in sequence, and at the time t, a transmission signal s of the ith periodtThe expression of (t, i) is:
wherein, A, f0,Respectively, amplitude of the transmitted signal, carrier frequency and initial phase, mu ═ B0T is the chirp rate, where B0Is the sweep bandwidth and T is the period of a chirped continuous wave. Considering a target having a radial distance r from the target vehicle in front of the radar and a radial velocity v when t is 0, the present invention is positive in the direction of approaching the radar along the radial velocity, so that the received signal s is positiverThe expression of (t, i) can be written as:
wherein A is0Is the amplitude of the received signal, τ -2 (r-vt)/c is the time delay due to the distance between the target and the radar, and c is the speed of light. The received signal and the original transmitted signal are mixed and an intermediate frequency signal, also called beat signal, is obtained by a low pass filter. At time t, expression s for the ith periodic beat signalIF(t, i) can be written as:
further, in step 2, consider that the millimeter wave vehicle radar system has NtRoot transmitting antenna, NrRoot receiving antenna, NtThe same frequency modulation continuous wave is transmitted by the root transmitting antenna in turn, and the total transmission period length T1=NtT, and constitute a virtual array. Transmitting N in one signal processing flow timesaA chirped continuous wave. In the ith transmission period, the echo complex signal of the mth transmission antenna transmission signal received by the kth receiving antenna is:
wherein, i ═ 1.., Nsa,m=1,...,Nt,k=1,...,Nr,A,A0,f0,The amplitude of the transmitted signal, the amplitude of the received signal, the carrier frequency and the initial phase, respectively, mu-B0T is the chirp rate, where B0Is the sweep bandwidth, T is the transmit period length, d is the antenna spacing, and θ is the azimuth angle at which the target is located.
For each transmitted signal yk,i(t) sampling with the number of sampling points Ns=fsT, wherein fsIs the sampling frequency of the DSP digital chip. Then during the i-th transmission period,echo complex signal sampling signal of mth transmitting antenna transmitting signal received by kth receiving antenna And j-th sample of the echo complex signal of the m-th transmitting antenna transmitting signal received by the k-th receiving antenna in the i-th transmitting period. In a signal processing flow period T2Therein, T2=Nsa·NtT, k-th receiving antenna receiving sampled signals of m-th transmitting antenna transmitting signalsComprises the following steps:
will be one transmission period T1The echo signals of the transmitting signals of different transmitting antennas received by the same receiving antenna are combined, and the received signal in the ith transmitting period isThe whole signal processing flow T2Temporally forming virtual arrays
Wherein, i ═ 1.., Nsa,T1=NtT, so the intermediate frequency signal data matrix is NtNr×Nsa×NsThe data structure is a three-dimensional data vector, a time dimension, namely a fast time dimension, a speed dimension, namely a slow time dimension, and a phase dimension caused by the multi-transmitting multi-receiving virtual array, and is shown in the attached drawing.
Further, in step 3, FFT is performed on the fast time dimension and the slow time dimension of the received data according to the definition in step 2 on the echo signal expression obtained in step 1.
Performing windowing FFT processing on the received signal along the fast time dimension to obtain the received signal of the kth receiving antenna in the ith transmitting periodFor example, the following steps are carried out:
wherein, i is 1sa,k=1,...,Nr,wqIs a window function of NSA column vector of x 1, symbol &representsthe Hadamard product of two vectors, i.e., the corresponding elements are multiplied, FFT (·) indicates an FFT operation on the signal,indicating that the kth receiving antenna receives the signal after finishing the window FFT along the fast time dimension in the ith transmitting period.
Assuming that there is a target with a distance r and a velocity v, after performing FFT on the fast time dimension, the target spectrum peak position is:
since the sawtooth sweep period T is very small, fr,v≈2B0r/cT. Thus, the fast time dimension may be equivalent to the distance dimension and the spectral cells may be equivalent to the distance cells.
Reception per antenna per transmission cyclePerforming windowing FFT on the signal to obtainNqFFTThe number of FFT points in the fast time dimension.
For YVFPerforming FFT in slow time dimension to receive data of the kth receiving antenna and the l-th frequency spectrum unitFor example, the following steps are carried out:
wherein k is 1r,l=1,...,Ns,wsIs a window function of NsaA column vector of x 1, and,indicating the kth receiving antenna, the l-th spectrum unit performs the received signal after the windowed FFT processing along the slow time dimension.
After FFT is performed on the slow time dimension, the target spectrum peak position is:
after the FFT in the slow time dimension, the position of the target spectrum peak is only related to the speed, so the slow time dimension can be regarded as the speed dimension.
Carrying out slow time dimension FFT on the received signals of different distance units of each receiving antenna to obtainNsFFTThe number of FFT points in the slow time dimension.
Consider the result ofr,vAnd fvA target vehicle exists in the ith distance dimension unit and the ith speed dimension unit, and the complex vector of the frequency spectrum unit where the target is located is takenAnd recording as Y:
1, N, whereqFFT,l=1,...,NsFFT。
Further, in the step 4, a uniform equidistant linear array is considered, the number of array elements is M, the array element interval is d, K far-field narrow-band signal sources are provided, where K is not more than M, the number of fast time-dimension sampling points is also called as N. The far-field narrow-band signal source model is as follows:
Y=AS+N
where Y is the received signal of the array M × N, A is the manifold matrix of the array M × K, and S is the transmitted signal vector of K × N, where each row of S is S in step 1IF(t, i) matrix form after sampling. When considering an additive white gaussian noise vector matrix with Ν being mxn and the snapshot number N being 1, Y is the complex vector Y of the spectrum unit where the target is located in step 3.
Consider the dimension of the beamforming matrix B to be M B, where B is equal to or less than M, and B is a normal orthogonal matrix:
BHB=I
in the application scenario of the vehicle-mounted millimeter wave radar system, the field of view (FOV) is a limited angular range, and a beam former covering 360 degrees does not need to be designed. Under the premise, the azimuth angle information of the vehicle to be detected is considered to be in a known angle range, and an optimized beam forming matrix is obtained by utilizing the range information of the azimuth angle.
In the case that there are two vehicles to be detected, in order to estimate the azimuth angles of the two vehicles more accurately, an optimized beam forming matrix needs to be designed. Optimized beamforming matrix BoptThe following conditions need to be satisfied: from a guide vector a (theta)1),a(θ2) And a (theta)m) The determined subspace is contained in BoptTheta in the subspace defined by the columns of1,θ2Is the azimuth of the two objects, and θmCan be defined as:
θm=(θ1+θ2)/2
in the application scene of the millimeter wave vehicle-mounted radar, the following method can be adopted to obtain the optimized beam forming matrix Bopt:
Βopt=[υ1 … υB]
Where a (θ) is the steering vector of the array, θaAnd thetabIs the boundary of the azimuthal range, uiIs Q∞B large eigenvalues (λ)1≥λ2≥…≥λB>λB+1≥λB+2≥…≥λM=σ2) Diagonal matrix, BETA, as main diagonal elementsoptIs as upsiloniIs a matrix of column vectors, i ═ 1, 2.
The complex vector Y of the received signal in the specific range-Doppler unit in step 3 is passed through a beam former, and the beam forming matrix is B in this stepoptThe received signal Y is converted from the array element domain to the beam domain by the beam former to obtain a beam domain received signal vector YB。
YB=Bopt HY
Further, in step 5, the beam domain receives the signal vector YBSample covariance matrix ofCan be expressed as:
where I is the identity matrix and T ═ Bopt HA is an array of beam domain equivalentsThe matrix of the column flow pattern, A being the matrix of steering vectors, Rs=E[SSH]For the autocovariance matrix of the transmitted signal, RBIs a covariance matrix, σ, of the beam domain of the received signal without taking into account noise2Variance of gaussian white noise.
Since the received signal length is finite, the sample covariance matrixIs an estimate of the covariance, and its computational expression in practical applications is usually written as:
wherein, YB(i) Is a beam domain received signal vector YBThe ith element of (1). Under the condition of single snapshot of the vehicle-mounted radar, only B data can be used, which seriously influences the accuracy of covariance matrix estimation and further influences the effectiveness of angle estimation. In the case of a single snapshot (N ═ 1), the beam domain received signal model may be modified according to the far field narrowband signal source model described in step 4 as:
YB=BHAS+BHN
wherein, YB=[Y1,Y2,…,YB]TIs a B × 1 column vector, YiIs YBB is the beamformer dimension.
If Y isBIs a wide stationary process, and the covariance matrix of the wide stationary process can be known according to the knowledge of the engineering matrix and the digital signal processingMust be a conjugate symmetric Toeplitz matrix, so the sample covariance matrix can be written as:
element r of upper ith diagonal lineiThe autocorrelation function for the spatial domain can be written as:
ri=E[Ym+iY* m]
wherein, Ym+iRepresents YBB-1, which is the case for a single snapshot, is reduced from the matrix to a vector, which can be seen as a random process if Y is presentBOr a traversal process, then we can replace the mathematical expectation of the sample with the average over time, so riCan be further written as:
in practice, we use the following formula to calculate the sample covariance matrix:
the above expression for estimating the covariance of the samples is actually a biased estimator,and riBetweenThe difference is as follows:
In order to improve the performance of the wave beam domain MUSIC angle measurement algorithm under the condition of single snapshot, we modifyTo ensure that it is an unbiased estimator, it is not difficult to find that after correction using the following equationIs constantly equal to 0, and newIs described in (1).
As can be seen from the above equation, assurance is providedIs physically realizable, it is necessary to ensure that the subscript of Y satisfies m + i ≦ B of 1 ≦ m. That is, for a certain fixed i, there are B-i data that can be used, and B-i averaging can be performed. For example for r0We can use B data to calculate, which is equivalent to doing B averages, and for rB-1Only 1 data can be used for estimation, which is equivalent to not doing any averaging, this pair of synergyVariance matrixThe estimation is disadvantageous.
So that r is actuallyiThe highest i in the number of B is usually not B-1, but is 2/3-4/5 of B. But doing so reduces the aperture size of the antenna array, and the reduced degrees of freedom also means that the number of signal sources that can be estimated is reduced. Therefore, the present invention also adopts the method of dividing sub-beams, each of which has dimension of L × L, i < th > beamthSub-beams RiiI.e. the covariance matrix of the original sampleFrom the ith row to the (i + L) th row, and from the ith column to the (i + L) th column. And under the condition of ensuring the full rank of each sub-beam, the accuracy of the covariance matrix of the sample is improved.
And finally, the sample covariance matrix expression used for the MUSIC algorithm angle measurement is as follows:
where P is the number of beamlets, and B + P-1 is satisfied, the beamlet division is schematically illustrated in fig. 2.
Further, in step 6, the covariance matrix of the roar-corrected sample obtained in step 5 is subjected to the above-mentioned processPerforming eigenvalue decomposition to obtain noise subspace
Since the signal and the noise are independent from each other, the sample covariance matrix can be decomposed into two parts related to the signal and the noise. To pairThe characteristic decomposition is carried out as follows:
in the formula, sigmasSo as to makeThe first K maximum eigenvalues lambda obtained by characteristic decomposition1≥λ2≥…≥λKDiagonal matrix, sigma, as main diagonal elementsoIs thatThe remaining L-K eigenvalues lambdaK+1≥λK+2≥…≥λL=σ2A diagonal matrix as its main diagonal elements, where L isK is the number of detected target vehicles, σ2Is the variance of the noise.Is determined by the first K maximum eigenvalues λ1≥λ2≥…≥λKThe subspace spanned by the corresponding feature vectors is the signal subspace, andis determined by a characteristic value lambdaK+1≥λK+2≥…≥λL=σ2The subspace spanned by the corresponding feature vectors is also the noise subspace.
In step 6, the proposed new MUSIC spatial spectrum function can be written as:
wherein b (θ) ═<BHa(θ)>LIs the first L elements of the steering vector of the beam domain,<·>Lrepresenting a column-taking vectorThe first L elements of (a) are operated on,is thatThe pseudo-inverse of (1). Ideally, it is assumed that there is an azimuth angle θiFor theta ∈ [0 DEG 360 DEG ]]Go through the traversal, when theta is equal to thetaiWhen the temperature of the water is higher than the set temperature,siis oriented at an angle thetaiCorresponding to the detection target vehicle iAnd decomposing the eigenvalue to obtain the eigenvalue. And in the ideal case the signal subspace and the noise subspace are orthogonal to each other, that is to say the steering vector in the signal subspace is also orthogonal to the noise subspace, that is to say:
because the steering vectors in the signal subspace are also not perfectly orthogonal to the noise subspace due to the presence of noise, the azimuthal angle estimate of the spatial spectrum estimate is actually estimated in terms of θ ∈ [0 ° 360 ° ]]By performing traversal, maximum-optimisation search, i.e. azimuth estimation of the detected target vehicleComprises the following steps:
wherein the content of the first and second substances,operation expression is obtainedThe value of θ corresponding to the maximum value is obtained.
Has the advantages that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
aiming at the characteristics of limited digital signal processing capability and high real-time requirement of a millimeter wave vehicle-mounted radar system, the optimized beam domain MUSIC algorithm is adopted to reduce the calculation complexity and the memory occupation. Aiming at the defects that the target resolution performance of the conventional MUSIC method is reduced under the conditions of low signal-to-noise ratio and small snapshot number and the engineering application under the non-ideal condition is difficult to adapt, the novel MUSIC estimator is provided, the information contained in a signal subspace, a noise subspace and a main characteristic value is utilized to the maximum extent, the advantages of high signal subspace processing robustness and high noise subspace processing estimation precision are combined, and the multi-target azimuth estimation performance is improved.
Drawings
FIG. 1 is a flow chart of the algorithm design of the present invention;
FIG. 2 is a schematic diagram of sub-beam splitting according to the present invention;
FIG. 3 is a graph of the detection success probability of the four MUSIC algorithms according to the embodiment 1 of the present invention varying with SNR;
FIG. 4 is a graph of the mean value of the detection angle errors of the four MUSIC algorithms varying with SNR in embodiment 1 of the present invention;
FIG. 5 is a graph of standard deviation of detection angle errors of four MUSIC algorithms varying with SNR in embodiment 1 of the present invention;
FIG. 6 is a graph of the angular resolution success probability of two MUSIC algorithms (MA3, MA4) as a function of SNR in example 2 of the present invention;
FIG. 7 is a graph of the mean value of the detection angle errors of two MUSIC algorithms (MA3, MA4) according to the embodiment of the present invention 2 along with the change of SNR;
FIG. 8 is a graph of standard deviation of detection angle errors of two MUSIC algorithms (MA3, MA4) according to the embodiment of the present invention 2 as a function of SNR;
FIG. 9 is a graph of the angular resolution success rate versus SNR for two MUSIC algorithms (MA3, MA4) at different angular intervals Δ θ in example 2 of the present invention;
FIG. 10 is a three-dimensional graphical representation of a block of radar data within a coherent processing interval in accordance with the present invention;
fig. 11 shows a possible azimuth distribution range of the vehicle to be detected in the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are intended to be purely exemplary of the invention and are not intended to limit its scope, as various equivalent modifications of the invention will become apparent to those skilled in the art after reading the present invention and fall within the scope of the appended claims.
In this section, we present some specific numerical results to further illustrate the effectiveness of the enhanced beam-space MUSIC algorithm proposed by the present invention. The superiority of the proposed enhanced beam space MUSIC algorithm over the traditional MUSIC algorithm under different conditions of a single target and two adjacent position targets is analyzed by the following two specific embodiments.
We compared the performance of the following four different MUSIC algorithms. For convenience of explanation, they are referred to as MA0(MUSIC algorithm 0), MA1(MUSIC algorithm 1), MA2(MUSIC algorithm 2), MA3(MUSIC algorithm 3), and MA4(MUSIC algorithm 4). MA0 is the classical MUSIC algorithm in the array element space. MA1 is the beam space MUSIC algorithm using DFT beamformers. MA2 is a beam former BoptThe beam space MUSIC algorithm of (1). MA3 is a method of using a beamformer and a sample covariance matrixIn order to maintain its Toeplitz structure in the case of single snapshots. MA4 is the enhanced beam space MUSIC algorithm proposed by the present invention. The specific differences between the five different algorithms are shown in table 2.
TABLE 2 comparison of differences between the five algorithms
Case 1: the performance of the single-signal source classical MUSIC algorithm is compared with the beam space MUSIC algorithm. In this case, we compare the performance of different beam space MUSIC algorithms with classical array element domain MUSIC in terms of angle measurement. The number of signal sources is K is 1. Boundary theta of angular rangeaAnd thetabAt 3 deg. and-3 deg., respectively. Target is at theta12.5. The fast beat number N is 1, and the array element number M is 8. We selected four different MUSIC algorithms (MA0, MA1, MA2 and MA4) for performance comparison.
As can be seen from the simulation results of fig. 3 to 5, the performance of the enhanced beam space MUSIC algorithm (MA4) proposed herein is optimal in terms of success probability, mean of deviation, and standard deviation among the four schemes. While the other two beamforming MUSIC algorithms are inferior to the classical element domain MUSIC algorithm (MA0) because the computational complexity is reduced at the cost of antenna aperture loss. Furthermore, the beam space MUSIC algorithm using DFT beamformer (MA1) has the worst angular performance among the four algorithms, which also demonstrates the optimized beamforming matrix B proposed by the present invention using a priori informationoptThe effectiveness of (c).
Case 2: performance comparisons of the single snapshot beam-space MUSIC algorithm (MA3, MA4) for two adjacent targets. First, we use the following two tables to illustrate the adverse effect of snapshot count reduction on signal source correlation in the case of two targets. The number of array elements M is 8, and the number of wave beams B is 6. As can be seen from table 3, in the case of large fast beats, the feature vector orthogonality of the noise subspace and the steering vector is significant, but in the case of small fast beats, the orthogonality between the obtained noise feature vector and the signal subspace deteriorates severely, resulting in inaccurate spectral estimation. It can also be seen from table 4 that by modifying the expression of the sample covariance matrix with MA3, maintaining the Toeplitz structure, the orthogonality of the feature vectors of the noise subspace and the steering vector is increased by an order of magnitude, which is very important for the subsequent DOA estimation.
TABLE 3 correlation analysis of MA0
Table 4 correlation analysis of MA0, MA3 (N ═ 1)
MA4 is further improved on the basis of MA 3. Therefore, MA4 can also guarantee orthogonality between the eigenvectors of the noise subspace and the steering vectors, while the performance of the angle measurement is further improved than that of MA 3. FIGS. 6-8 show the enhanced beam space MUSIC algorithm proposed by the present invention at two targets, theta1=5°,θ2The resolution is higher in the case of-5 °, the mean deviation is smaller and the standard deviation is smaller. Figure 9 shows a comparison of the resolution probabilities of two MUSIC algorithms at different angular intervals. When the angular interval Δ θ is the same, MA4 has a higher resolution probability at the same SNR throughout. It is worth noting that when both targets approach Δ θ — 4 °, MA3 is no longer inactive and MA4 is still active. Let us assume that the number of array elements M is 8, the number of beams B is 6, and the number of fast beats N is 1.
Claims (7)
1. A dimension reduction subspace angle measurement method suitable for the millimeter wave vehicle-mounted radar target detection field is characterized by comprising the following steps:
step 1: establishing a space spectrum estimation mathematical model of the millimeter wave vehicle-mounted radar system to obtain expressions of transmitting signals and receiving signals;
step 2: expanding and establishing a three-dimensional data structure of the intermediate frequency signal after mixing the received signal under the condition of multiple receiving and transmitting on the basis of the step 1 by utilizing the number and the layout of transmitting antennas and receiving antennas required by a millimeter wave vehicle-mounted radar system;
and step 3: performing FFT on the three-dimensional received signal data obtained in the step 2 in a fast time dimension and a slow time dimension respectively to obtain the distance and speed information of the detected target vehicle and a received signal data vector Y of the antenna array on the corresponding distance Doppler unit;
and 4, step 4: calculating an optimized beam forming matrix B by using the azimuth angle range information of the vehicle to be detected by the vehicle radaroptConverting the received signal data vector Y of the antenna array on the specific range Doppler unit from the array element domain to the beam domain to obtain the received signal vector Y of the beam domainB;
And 5: using the beam domain received signal matrix Y obtained in step 4BComputing a sample covariance matrixAnd toCorrecting to obtain a corrected sample covariance matrix
2. The method for measuring the angle of the dimensionality reduced subspace, which is suitable for the millimeter wave vehicle-mounted radar target detection field according to claim 1, is characterized in that in the step 1, a space spectrum estimation mathematical model of a millimeter wave vehicle-mounted radar system is established, and expressions of a transmitting signal and a receiving signal are obtained, wherein the method comprises the following steps:
the wave form of the transmitting signal of the millimeter wave vehicle-mounted radar system is a group of carrier frequency f0In the transmitting period, a linear frequency modulation continuous wave signal with a certain sweep frequency bandwidth is transmitted by a plurality of transmitting antennas in sequence in a time-sharing manner, and at the time t, the transmitting signal s of the ith periodtThe expression of (t, i) is:
wherein, A, f0,Respectively, amplitude of the transmitted signal, carrier frequency and initial phase, mu ═ B0T is the chirp rate, where B0Is sweep frequency bandwidth, T is a period of the linear frequency modulation continuous wave, when T is 0, the radial distance between the front of the radar and the target vehicle is r, the target with radial velocity v is positive along the direction that the radial velocity approaches the radar, so the received signal srThe expression of (t, i) can be written as:
wherein A is0Is the amplitude of the received signal, τ is 2(r-vt)/c is the time delay caused by the distance between the target and the radar, c is the speed of light, the received signal and the original transmitted signal are mixed, and an intermediate frequency signal, also called beat signal, is obtained by a low pass filterNumber, at time t, expression s for the ith periodic beat signalIF(t, i) can be written as:
3. the method for measuring the angle in the dimension-reduced subspace suitable for the millimeter wave vehicle radar target detection field according to claim 2, wherein in the step (2), a three-dimensional data structure of the intermediate frequency signal after the received signal is mixed under the condition of multiple transceiving is expanded and established on the basis of the step (1), and the method comprises the following steps:
(2.1) the millimeter wave vehicular radar system has NtRoot transmitting antenna, NrRoot receiving antenna, NtThe same frequency modulation continuous wave is transmitted by the root transmitting antenna in turn, and the total transmission period length T1=NtT, and forming a virtual array, transmitting N in one signal processing flow timesaIn the ith transmitting cycle, the echo complex signal of the mth transmitting antenna transmitting signal received by the kth receiving antenna is:
wherein, i ═ 1.., Nsa,m=1,...,Nt,k=1,...,Nr,A,A0,f0,The amplitude of the transmitted signal, the amplitude of the received signal, the carrier frequency and the initial phase, respectively, mu-B0T is the chirp rate, where B0The method comprises the following steps that frequency sweep bandwidth is adopted, T is the length of a transmitting period, d is the distance between antennas, and theta is the size of an azimuth angle of a target;
(2.2) for each transmitted signal yk,i(t) sampling with the number of sampling points Ns=fsT, wherein,fsThe sampling frequency of the DSP digital chip, in the ith transmitting period, the echo complex signal sampling signal of the m transmitting antenna transmitting signal received by the k receiving antenna The j-th sample of the echo complex signal of the m-th transmitting antenna transmitting signal received by the k-th receiving antenna in the i-th transmitting period is shown in a signal processing flow period T2Therein, T2=Nsa·NtT, k-th receiving antenna receiving sampled signals of m-th transmitting antenna transmitting signalsComprises the following steps:
will be one transmission period T1The echo signals of the transmitting signals of different transmitting antennas received by the same receiving antenna are combined, and the received signal in the ith transmitting period isThe whole signal processing flow T2Temporally forming virtual arrays
Wherein, i ═ 1.., Nsa,T1=NtT, so the intermediate frequency signal data matrix is NtNr×Nsa×NsWhich is a three-dimensional data vector.
4. The dimension-reduced subspace angle measurement method suitable for the millimeter wave vehicle-mounted radar target detection field according to claim 3, wherein in the step (3), FFT is respectively performed on the three-dimensional received signal data obtained in the step (2) in the fast time dimension and the slow time dimension to obtain the distance and speed information of the detected target vehicle and the received signal data vector Y of the antenna array on the corresponding range-doppler unit, and the method is as follows:
(3.1) carrying out windowing FFT processing on the received signal along the fast time dimension, and setting the received signal of the kth receiving antenna in the ith transmitting periodPerforming windowing FFT processing:
wherein, i is 1sa,k=1,...,Nr,wqIs a window function of NSA column vector of x 1, symbol &representsthe Hadamard product of two vectors, i.e., the corresponding elements are multiplied, FFT (·) indicates an FFT operation on the signal,indicating that the kth receiving antenna receives signals after finishing the window FFT along the fast time dimension in the ith transmitting period;
(3.2) assuming that a target vehicle with a radial distance r and a radial speed v from the radar exists, after FFT is carried out on the fast time dimension, the peak position of a target frequency spectrum is as follows:
let fr,v=2B0r/cT, the fast time dimension can be equivalent to a distance dimension, and the frequency spectrum unit can be equivalent to a distance unit;
(3.3) carrying out windowing FFT (fast Fourier transform) on the received signal of each antenna in each transmission period to obtainNqFFTFast time dimension FFT points;
(3.4) for YVFPerforming FFT in slow time dimension, setting kth receiving antenna and the data of the first frequency spectrum unitThe FFT processing is performed as follows:
wherein k is 1r,l=1,...,Ns,wsIs a window function of NsaA column vector of x 1, and,representing the kth receiving antenna, and the l frequency spectrum unit receives signals after finishing the window FFT processing along the slow time dimension;
after FFT is performed on the slow time dimension, the target spectrum peak position is:
after FFT of the slow time dimension, the position of the target frequency spectrum peak is only related to the speed, and the slow time dimension is regarded as the speed dimension;
(3.5) for each receive antenna,the receiving signals of different distance units are subjected to slow time dimension FFT to obtainNsFFTThe number of FFT points in the slow time dimension;
(3.6) from fr,vAnd fvA target vehicle exists in the ith distance dimension unit and the ith speed dimension unit, and the complex vector of the frequency spectrum unit where the target is located is takenAnd recording as Y:
1, N, whereqFFT,l=1,...,NsFFT。
5. The method for measuring the angle of the dimensionality-reduced subspace, which is suitable for the millimeter wave vehicle-mounted radar target detection field, according to the claim 4, wherein in the step (4), the optimized beam forming matrix B is calculated by using the information of the azimuth angle range where the vehicle to be detected by the vehicle-mounted radar is locatedoptConverting the received signal data vector Y of the antenna array on the specific range Doppler unit from the array element domain to the beam domain to obtain the received signal vector Y of the beam domainBThe method comprises the following steps:
(4.1) a uniform equidistant linear array is arranged, the number of array elements is M, the spacing between the array elements is d, and a far-field narrow-band signal source model is as follows:
Y=AS+N
wherein Y is the receiving signal of the array M × N, A is the array manifold matrix of M × K, and S is the transmitting signal vector of K × N, where each row of S is S in step (1)IF(t, i) in a matrix form after sampling, wherein N is an additive white gaussian noise vector matrix of mxn, K is the number of far-field narrow-band signal sources, K is less than or equal to M, the number of fast time-dimension sampling points is also called as fast beat number N, and when the fast beat number N is 1, Y is the frequency of the target in step (3)Complex vector Y of spectral unit;
(4.2) the dimension of the beamforming matrix B is M B, and B is less than or equal to M, while B is an orthonormal matrix:
BHB=I
(4.3) in the presence of two vehicles to be detected, in order to estimate the horizontal azimuth angle of the two vehicles relative to the radar, a beam forming matrix B is designed, and the designed beam forming matrix B is recorded as BoptThe matrix BoptThe following conditions need to be satisfied: from a guide vector a (theta)1),a(θ2) And a (theta)m) The determined subspace is contained in BoptIn the subspace defined by the columns, the steering vector is calculated using the following formula:
where d is the antenna spacing, λ is the carrier wavelength, θ1,θ2Is the azimuth of the two objects, and θmCan be defined as:
θm=(θ1+θ2)/2
(4.4) obtaining the beamforming matrix B by the following methodopt:
Βopt=[υ1…υB]
Where a (θ) is the steering vector of the array, θaAnd thetabIs the boundary of the azimuth range, λ1≥λ2≥…≥λBIs Q∞The first B large eigenvalues, upsilon, obtained by eigenvalue decompositioniIs the characteristic vector, BETA, corresponding to the B large characteristic valuesoptIs as upsiloniA matrix of column vectors, i ═ 1,2,. B;
(4.5) receiving signals on the specific range-Doppler cell described in step (3)The complex vector Y is passed through a beam forming matrix, which is B in this stepoptConverting the received signal Y from the array element domain to the beam domain through the beam forming matrix to obtain a beam domain received signal vector YB:
YB=Bopt HY。
6. The method for measuring angle in subspace with reduced dimension suitable for the millimeter wave vehicle-mounted radar target detection field as claimed in claim 5, wherein in step (5), the beam domain received signal matrix Y obtained in step 4 is usedBComputing a sample covariance matrixAnd toCorrecting to obtain a corrected sample covariance matrixThe method comprises the following steps:
(5.1) when the snapshot number N is equal to 1, ensuring the covariance matrix of the sampleThe estimate of (d) is unbiased, which can be written as:
in the formula (I), the compound is shown in the specification,is thatThe elements of the upper ith diagonal can be written as:
where B is the dimension of the beamforming matrix and YmIs a received signal vector YBM-th element of (a), m-0, 1.
(5.2) adopting a sub-beam averaging method,to representThe ith sub-beam of (1), the ith sub-beam data being the covariance matrix of the original sampleFrom row i to row i + L and column i to column i + L, and then P sub-beam data blocks for i 1,2Averaging to obtain a final corrected sample correction variance matrix
Wherein, P is the number of sub-beams, and B is L + P-1.
7. The method for measuring angle in subspace of millimeter wave vehicle-mounted radar in object detection field according to claim 6, wherein in step (6), the corrected sample covariance matrix of step 5 is appliedPerforming eigenvalue decomposition to obtain noise subspaceBy usingAndestablishing a spatial spectral function feMUSICPerforming spectral peak search to obtain an incoming wave direction estimateThe method comprises the following steps:
(6.1) since the signal and noise are independent of each other, the sample covariance matrix can be decomposed into two parts related to the signal and noise, and the two parts are pairedThe characteristic decomposition is carried out as follows:
in the formula, sigmasSo as to makeThe first K maximum eigenvalues lambda obtained by characteristic decomposition1≥λ2≥…≥λKAs a diagonal matrix of major diagonal elements, ∑ o isThe remaining L-K eigenvalues lambdaK+1≥λK+2≥…≥λL=σ2A diagonal matrix as its main diagonal elements, where L isK is the number of detected target vehicles, σ2Is the variance of the noise and is,is determined by the first K maximum eigenvalues λ1≥λ2≥…≥λKThe subspace spanned by the corresponding feature vectors, i.e. the signal subspace, ofIs determined by a characteristic value lambdaK+1≥λK+2≥…≥λL=σ2A subspace spanned by the corresponding feature vectors is also called a noise subspace;
(6.2) the built MUSIC spatial spectrum function is as follows:
wherein b (θ) ═<BHa(θ)>LIs the first L elements of the steering vector of the beam domain,<·>Lrepresenting the first L element operations of the column vector,is thatUnder ideal conditions, an azimuth angle theta is assumed to existiFor theta ∈ [0 DEG 360 DEG ]]Go through the traversal, when theta is equal to thetaiWhen the temperature of the water is higher than the set temperature,siis oriented at an angle thetaiCorresponding to the detection target vehicle iEigenvalue decomposition instituteObtaining a characteristic value;
(6.3) in an ideal state, θ ═ θiThe time signal subspace and the noise subspace are mutually orthogonal, i.e. the steering vectors in the signal subspace are also orthogonal to the noise subspace, i.e.:
because the steering vectors in the signal subspace are also not perfectly orthogonal to the noise subspace due to the presence of noise, the azimuthal angle estimate of the spatial spectrum estimate is actually estimated in terms of θ ∈ [0 ° 360 ° ]]Traversing, and optimally searching the maximum value to realize the azimuth angle theta of the targetiIs estimated byI.e. azimuth estimation of detected target vehiclesComprises the following steps:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910238433.3A CN109975807B (en) | 2019-03-27 | 2019-03-27 | Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910238433.3A CN109975807B (en) | 2019-03-27 | 2019-03-27 | Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109975807A CN109975807A (en) | 2019-07-05 |
CN109975807B true CN109975807B (en) | 2022-03-18 |
Family
ID=67080953
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910238433.3A Active CN109975807B (en) | 2019-03-27 | 2019-03-27 | Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109975807B (en) |
Families Citing this family (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110515066B (en) * | 2019-08-20 | 2023-03-24 | 立晟智能科技(成都)有限公司 | Vehicle-mounted millimeter wave radar and target height measuring method thereof |
CN110632586B (en) * | 2019-10-10 | 2022-05-24 | 南京凌云科技发展有限公司 | Road vehicle low-computation monitoring method based on rapid FMCW radar |
CN111044996A (en) * | 2019-12-20 | 2020-04-21 | 南京航空航天大学 | LFMCW radar target detection method based on dimension reduction approximate message transfer |
CN111175753B (en) * | 2020-02-24 | 2021-08-17 | 四川九洲电器集团有限责任公司 | Vehicle-mounted anti-collision radar wave-splitting target detection method |
CN113391257B (en) * | 2020-03-13 | 2023-04-18 | 光宝科技新加坡私人有限公司 | Computing device for object angle estimation and object angle estimation method |
CN111537966B (en) * | 2020-04-28 | 2022-06-10 | 东南大学 | Array antenna error correction method suitable for millimeter wave vehicle-mounted radar field |
CN111521968B (en) * | 2020-05-22 | 2022-05-20 | 南京理工大学 | Underdetermined DOA estimation method based on target space diversity |
CN112711013B (en) * | 2020-12-14 | 2022-10-21 | 中国船舶重工集团公司第七一五研究所 | Rapid self-adaptive beam forming method based on block matrix |
CN116569062A (en) * | 2021-02-19 | 2023-08-08 | 华为技术有限公司 | Positioning method and device |
CN113030868A (en) * | 2021-03-29 | 2021-06-25 | 长沙莫之比智能科技有限公司 | Millimeter wave radar angle super-resolution method |
CN114609626A (en) * | 2022-03-16 | 2022-06-10 | 电子科技大学 | Multi-target detection method for vehicle-mounted millimeter wave radar |
CN114428236B (en) * | 2022-04-06 | 2022-07-01 | 南京楚航科技有限公司 | Vehicle-mounted millimeter wave radar angle confidence degree estimation method |
CN115113157B (en) * | 2022-08-29 | 2022-11-22 | 成都瑞达物联科技有限公司 | Beam pointing calibration method based on vehicle-road cooperative radar |
CN116819509B (en) * | 2023-08-28 | 2023-11-07 | 烟台初心航空科技有限公司 | Radar positioning and ranging method based on spread spectrum time domain reflection |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH10142331A (en) * | 1996-11-14 | 1998-05-29 | Komatsu Ltd | Millimetric wave radar-loaded vehicle |
CN103529444A (en) * | 2013-09-27 | 2014-01-22 | 安徽师范大学 | Vehicle-mounted millimeter-wave radar moving target recognizer and recognition method |
CN105487070A (en) * | 2014-10-06 | 2016-04-13 | 日本电产艾莱希斯株式会社 | Radar system, radar signal processing device, vehicle driving control device and method and computer program |
CN106842114A (en) * | 2016-12-29 | 2017-06-13 | 西安电子科技大学 | Target direction of arrival acquisition methods based on root MUSIC algorithms |
CN107688178A (en) * | 2017-08-25 | 2018-02-13 | 上海通趣科技有限公司 | A kind of sawtooth waveforms ranging and range rate method based on 77GHz millimetre-wave radars |
CN109358322A (en) * | 2018-10-25 | 2019-02-19 | 森思泰克河北科技有限公司 | Forward direction target detection radar and method |
-
2019
- 2019-03-27 CN CN201910238433.3A patent/CN109975807B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH10142331A (en) * | 1996-11-14 | 1998-05-29 | Komatsu Ltd | Millimetric wave radar-loaded vehicle |
CN103529444A (en) * | 2013-09-27 | 2014-01-22 | 安徽师范大学 | Vehicle-mounted millimeter-wave radar moving target recognizer and recognition method |
CN105487070A (en) * | 2014-10-06 | 2016-04-13 | 日本电产艾莱希斯株式会社 | Radar system, radar signal processing device, vehicle driving control device and method and computer program |
CN106842114A (en) * | 2016-12-29 | 2017-06-13 | 西安电子科技大学 | Target direction of arrival acquisition methods based on root MUSIC algorithms |
CN107688178A (en) * | 2017-08-25 | 2018-02-13 | 上海通趣科技有限公司 | A kind of sawtooth waveforms ranging and range rate method based on 77GHz millimetre-wave radars |
CN109358322A (en) * | 2018-10-25 | 2019-02-19 | 森思泰克河北科技有限公司 | Forward direction target detection radar and method |
Non-Patent Citations (1)
Title |
---|
基于遗传算法的车载雷达天线阵列布局优化方法;邹乐 等;《微波学报》;20080731;第34卷;第155-158页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109975807A (en) | 2019-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109975807B (en) | Dimension reduction subspace angle measurement method suitable for millimeter wave vehicle-mounted radar | |
CN111308437B (en) | Entropy-solving and speed-ambiguity-solving method for millimeter wave MIMO traffic radar | |
CN111537966B (en) | Array antenna error correction method suitable for millimeter wave vehicle-mounted radar field | |
CN110133631B (en) | Fuzzy function-based frequency control array MIMO radar target positioning method | |
CN110673086A (en) | Two-dimensional angle super-resolution method based on digital array radar | |
CN109765521B (en) | Beam domain imaging method based on subarray division | |
CN111007503B (en) | Moving target focusing and positioning method and system based on frequency spectrum accurate positioning | |
CN114895260A (en) | Echo separation method for pitching space-time coding space-borne SAR system | |
CN111521968B (en) | Underdetermined DOA estimation method based on target space diversity | |
CN108828504B (en) | MIMO radar target direction fast estimation method based on partial correlation waveform | |
Gao et al. | Frequency diverse array MIMO radar adaptive beamforming with range-dependent interference suppression in target localization | |
CN112180339A (en) | Radar echo signal accurate direction finding method based on sparse processing | |
CN108828586B (en) | Bistatic MIMO radar angle measurement optimization method based on beam domain | |
US11754671B2 (en) | Incoming wave count estimation apparatus and incoming wave count incoming direction estimation apparatus | |
CN110196417A (en) | The bistatic MIMO radar angle estimating method concentrated based on emitted energy | |
CN111175745B (en) | Moving target three-dimensional imaging method based on state space balance method | |
CN111368256A (en) | Single snapshot direction finding method based on uniform circular array | |
CN110579737A (en) | Sparse array-based MIMO radar broadband DOA calculation method in clutter environment | |
CN114779199A (en) | Adaptive cross coupling compensation and clutter suppression method for end-fire array airborne radar | |
CN114355309A (en) | Target angle and distance positioning method and device, radar and storage medium | |
CN115015925A (en) | Airborne array radar super-resolution forward-looking imaging method and device based on improved matching pursuit | |
CN112666558B (en) | Low-complexity MUSIC direction finding method and device suitable for automobile FMCW radar | |
WO2021196165A1 (en) | Frequency analysis method, device and radar | |
CN110231590B (en) | Array target angle measurement method based on DFT (discrete Fourier transform) | |
Schurwanz et al. | Compressive sensing techniques applied to a semi-circular mmWave MIMO array |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |