An adaptive constrained orthogonal matching pursuit algorithm
By using an adaptive constrained orthogonal matching tracking algorithm, the intermediate frequency signal is converted from the time domain to the frequency domain. Combined with compressed sensing technology, this solves the positioning accuracy problem in multi-target moving scenarios and achieves high-precision super-resolution positioning, which is suitable for industrial production, smart homes and autonomous driving.
Patent Information
- Application Number
- CN202410370154.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-29
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2044-03-29
AI Technical Summary
Existing radar positioning technology struggles to achieve high-precision super-resolution positioning in multi-target moving scenarios. In particular, when multiple target signals are superimposed, traditional orthogonal matching algorithms are highly dependent on sparsity, making it difficult to select iteration termination conditions, resulting in severe noise interference and affecting positioning accuracy.
An adaptive constrained orthogonal matching pursuit algorithm is adopted. By converting the intermediate frequency signal from the time domain to the frequency domain and combining it with compressed sensing technology, the sparse dependency is optimized, a positioning model is established, the distance and angle relationship between the object and the antenna is calculated, the iteration termination condition is optimized, and the positioning accuracy is improved.
It significantly improves object positioning accuracy in multi-target moving scenarios, and is applicable to fields such as industrial production, smart homes and autonomous driving, providing higher positioning reliability and accuracy.
Smart Images

Figure CN118091590B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of signal processing technology, specifically relating to a radar positioning and tracking technology, and more particularly to an adaptive constrained orthogonal matching tracking algorithm. Background Technology
[0002] Mobile scene positioning technology has irreplaceable practical value in the fields of rapid object detection, intelligent assisted driving, and autonomous driving. In daily life, work, or factory production, accurate target positioning enables automated capture, material handling, and a series of other operations. In transportation, precise target positioning provides real-time information about the surrounding environment.
[0003] Frequency-modulated continuous wave (FM-CHW) radar achieves object localization by transmitting signals with linearly increasing frequencies and receiving the reflected signals. Existing localization technologies struggle to meet the demands of multi-target moving scenarios, necessitating improvements in accuracy and reliability. The localization problem in multi-target moving scenarios is complex, requiring super-resolution localization. Maximizing object localization accuracy when multiple object signals are superimposed presents significant challenges.
[0004] Current resolution techniques involve mixing the echo signal with the transmitted signal to obtain an intermediate frequency (IF) signal. This IF signal is then windowed and filtered before being subjected to a Fast Fourier Transform (FFT) to decompose it into spectra with different separation peaks, enabling angle and range measurements, thus achieving super-resolution. Based on smoothed data, this method can obtain super-resolution positioning results.
[0005] With the addition of noisy data, the data perturbation is significant, with numerous local peaks, and the signal delay and angle are not integers of the number of Fourier transform points. The overlapping of signals from multiple targets leads to separation bias, resulting in unsatisfactory Fourier transform performance.
[0006] Traditional signal processing is based on Nyquist sampling, requiring digital systems to have a sampling rate more than twice the bandwidth of analog signals. In practical engineering, the sampling rate is much higher than twice. To save storage and computing resources, compressed processing of acquired data has emerged, leading to the development of compressed sensing, or CS, a sampling and processing technique for sparse signals.
[0007] The core problem of compressed sensing is signal reconstruction. A commonly used method is the orthogonal matching pursuit algorithm, or OMP for short. Traditional OMP algorithms are quite sensitive to signal sparsity K. When K is too large, the separation peaks are treated as clutter signals and masked; when K is too small, high-frequency clutter signals are treated as separation peaks, making it difficult to select the iteration termination condition. Summary of the Invention
[0008] The purpose of this invention is to address the technical problem of noise in traditional OMP (Optical Positioning Mechanism) by designing a super-resolution algorithm using intermediate frequency (IF) signals within one chirp cycle for multi-target movement scenarios. To address the sparsity of noise-overlapping signals, the IF signal is converted from a time-domain signal to a frequency-domain signal to obtain frequency-domain information and establish a positioning model. Based on the formula between frequency and distance, the distance and angle relationship between the object and the antenna are calculated. For traditional orthogonal matching algorithms, an adaptive orthogonal matching algorithm is used to optimize the iteration termination condition caused by high sparsity dependence. When multiple object signals are superimposed, compressed sensing technology is used to achieve super-resolution positioning and improve the object positioning accuracy.
[0009] To achieve the above objectives, the technical solution adopted by this invention is: an adaptive constrained orthogonal matching pursuit algorithm, comprising the following steps:
[0010] Step 1: After the radar antenna transmits a pulse signal, it performs down-conversion reception at the receiving end to obtain the intermediate frequency signal;
[0011] Step 2: Process the noisy intermediate frequency signal based on the adaptive orthogonal matching algorithm, and convert the intermediate frequency signal from the time domain signal to the frequency domain signal;
[0012] Step 3: Based on the frequency domain signal obtained in Step 2, calculate the position of the object in space and its relative angle to the antenna to achieve super-resolution.
[0013] In step 1: the signal S is transmitted. TX (t) After the antenna transmits, it will touch a space object and transmit again, thus obtaining the echo signal S. RX (t), after mixing and filtering the two, the intermediate frequency signal is obtained, specifically:
[0014] Let the transmitted signal S TX The expression for (t) is The signal frequency is f(t) = f0 + γt, where γ represents the frequency modulation slope, and the time t ranges from 0 to t to T. The FMCW signal with a duration of T is called a frequency modulation signal.
[0015] Let the echo signal S RX The expression for (t) is Where τ = 2R0 / c, τ represents the echo delay, R0 represents the target distance, and c represents the speed of light. Within one frequency modulation cycle, the transmitted signal and the echo signal overlap in time [τ, T]. The expression for the output intermediate frequency signal after the mixer is:
[0016]
[0017] When processing noisy intermediate frequency signals, the first step is to select test samples x and construct a training dictionary A. This is done by applying different radar signal strengths using the formula...
[0018] Select test sample x, where m represents the number of sampling points, N represents the number of radars, and Z_nosiy N Let i represent the noisy intermediate frequency signal of the Nth radar, where i represents an integer from 1 to m.
[0019] Construct a training dictionary using formulas Where Num represents an integer from 0 to 255, expressed as Num = [0, 1, 2, ..., 255]; sin represents the sine trigonometric function, c represents the speed of light, f0 represents the radar frequency; D represents the distance between radars, expressed as D = L / (N-1), where L represents the radar aperture; Ang is the deflection angle, expressed as Ang = linspace(Angle min ,Angle max ), where linespace(·) represents equal division, and angle min Indicates the minimum deflection angle, Angle max This indicates the maximum deflection angle.
[0020] Step 2 specifically includes:
[0021] Step 2.1: Input the test sample x and the training dictionary A, as well as the sparsity K;
[0022] Step 2.2: Let the initial residual r0 = x, the initial index set Λ0 = φ, the sensing matrix B0 = φ, and the iteration number t = 1, and initialize the signal;
[0023] Step 2.3: Use formula λ t =arg(max(| <r t-1 ,b j >|)),j=1,2,..,N calculate the index, according to the index λ t Searching, max(·) means taking the maximum value;
[0024] Step 2.4: The index set Λ for the t-th iteration t =Λ t-1 ∪λ t The sensing matrix B in the t-th iteration t =B t-1 ∪b λt ;
[0025] Step 2.5: Use the formula Calculate x = B t θ t The least squares solution;
[0026] Step 2.6: Use the formula Update residual r t ;
[0027] Step 2.7: Adaptively adjust the sparsity K using the residual norm difference; if t > 1, then use the formula Calculate the residual of the previous generation, if ||r t-1 -r t ||<||r t || t / 2 Then let t = t-1;
[0028] Step 2.8: If the iteration condition is met, then estimate the sparse representation coefficients of the output signal. Otherwise, return to step three and iterate again;
[0029] Where, r t Let φ represent the residual, t represent the number of iterations, φ represent the empty set, K represent the sparsity, B represent the sensing matrix, and Λ represent the sparseness. t Let λ represent the set of indices (column numbers) for t iterations. t b represents the index (column number) found in the t-th iteration. j Represents the j-th column of matrix B. t Indicates by index Λ t The set of columns of the selected matrix B, θ t It is a column vector, the symbol ∪ represents the union of sets, <·,·> represents the dot product of vectors, arg(min(.)) represents the minimum value of the expression, and arg(max(.)) represents the maximum value of the expression.
[0030] In step 3:
[0031] The distance R0 between the target and the antenna is expressed by the formula: Where f IF The expression is X represents the sampling point corresponding to the peak in the spectrum, n represents the number of sampling points, and F S Indicates the sampling frequency.
[0032] Assume that the distances between the same target detected by antennas 1 and 2 and the antennas are R1 and R2, respectively. Establish a coordinate system with the intersection of two circles with radii R1 and R2 as the target, connecting antennas 1 and 2 as the x-axis, and arbitrarily choosing a point on the line connecting them as the origin. Let the distance from antenna 1 to the origin be d. a The distance from antenna 2 to the origin is d. b The distance between antenna one and antenna two is d. 12 The angle between the line connecting antenna one and the target and the x-axis is θ1, expressed as: The angle between the line connecting antenna 2 and the target and the x-axis is θ2, expressed as: The target's ordinate is R1 sinθ1 or R2sinθ2, and its abscissa is R1cosθ1-d. a or d a -R2cosθ2.
[0033] Compared with the prior art, the technical solution adopted in this invention has the following beneficial effects:
[0034] 1. This invention addresses the sparsity of noise superimposed signals by employing compressed sensing technology, thereby improving the positioning accuracy of objects.
[0035] 2. This invention addresses the optimization problem of iteration termination conditions caused by the high sparsity dependence of traditional orthogonal matching algorithms. Based on the traditional matching algorithm, an adaptive orthogonal matching algorithm is designed to overcome the problem of high sparsity dependence of orthogonal matching algorithms, so as to ensure the optimality of iteration and achieve better matching.
[0036] 3. The algorithm proposed in this invention can provide object positioning accuracy in mobile scenarios and has irreplaceable practical value in fields such as industrial production, smart homes, and autonomous driving. Attached Figure Description
[0037] Figure 1 This is a flowchart illustrating an adaptive constraint orthogonal matching tracking algorithm in this embodiment.
[0038] Figure 2 This is a schematic diagram of the intermediate frequency signal generation process in an adaptive constrained orthogonal matching pursuit algorithm of this embodiment.
[0039] Figure 3 This is a schematic diagram of the overlapping of intermediate frequency signals of multiple targets in this embodiment.
[0040] Figure 4 This is a schematic diagram showing the positional relationship of the intersection points of the detection distances of different antenna ports in this embodiment.
[0041] Figure 5 This is a schematic diagram of the spatial positional relationship of the target in this embodiment.
[0042] Figure 6 This is a distance map between the noisy radar signal and the detected target in this embodiment.
[0043] Figure 7 This is a schematic diagram comparing the Fourier transform with the original OMP sparse reconstructed signal in this embodiment.
[0044] Figure 8 This is a schematic diagram of the OMP residual norm versus signal sparsity curves in this embodiment.
[0045] Figure 9This is a schematic diagram comparing the Fourier transform and the optimized OMP reconstructed signal in this embodiment.
[0046] Figure 10 This is a schematic diagram of the spatial positional relationship of multi-target detection in this embodiment.
[0047] Figure 11 This is a schematic diagram of the positional relationship of the target in polar coordinates in this embodiment.
[0048] Figure 12 This is a schematic diagram of the positional relationship of the target in two-dimensional coordinates in this embodiment. Detailed Implementation
[0049] The technical solution of the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments.
[0050] like Figure 1 As shown, this embodiment illustrates a flowchart of an adaptive constrained orthogonal matching pursuit algorithm, which includes the following steps:
[0051] Step 1: After the radar antenna transmits a pulse signal, it performs down-conversion reception at the receiving end to obtain the intermediate frequency signal;
[0052] Step 2: Process the noisy intermediate frequency signal based on the adaptive orthogonal matching algorithm, and convert the intermediate frequency signal from the time domain signal to the frequency domain signal;
[0053] Step 3: Based on the frequency domain signal obtained in Step 2, calculate the position of the object in space and its relative angle to the antenna to achieve super-resolution.
[0054] Transmit signal S TX (t) After the antenna transmits, and after it touches a space object and transmits again, the echo signal S is obtained. RX (t), the two are mixed and filtered to obtain the intermediate frequency signal. The detailed process is as follows: Figure 2 As shown.
[0055] set up Figure 2 Transmitted signal S TX The expression for (t) is The signal frequency is f(t) = f0 + γt, where γ represents the frequency modulation slope, and the time t ranges from 0 to t to T. The FMCW signal with a duration of T is called a frequency modulation signal.
[0056] Let the echo signal S RX The expression for (t) is Where R0 represents the target distance, τ represents the echo delay, and c represents the speed of light, then τ = 2R0 / c. Within one frequency modulation cycle, the transmitted signal and the echo signal overlap in time [τ, T]. The expression for the output intermediate frequency signal after the mixer is:
[0057] Because the locations of multiple detection targets are different, the echo delay of each reflected signal is also different, so it can be converted into multiple intermediate frequency signals with different constant frequencies, such as... Figure 3 As shown. Simultaneously, multiple antennas will receive signals reflected from the same target. This embodiment uses a millimeter-wave radar with 86 antenna elements to collect indoor detection data, operating at a frequency of 76-81 GHz. It completes intermediate frequency data acquisition by receiving reflected signals and performing frequency mixing.
[0058] This embodiment processes noisy intermediate frequency (IF) signals using an adaptive orthogonal matching algorithm, converting the IF signal from a time-domain signal to a frequency-domain signal. Based on the established positioning model, it calculates the position of objects in space and their relative antenna angles. The principle of the adaptive orthogonal matching algorithm is as follows:
[0059] When processing radar signals with clutter, the first step is to select test samples x and construct a training dictionary A. This is done by applying different radar signal strengths using formulas... Select test sample x, where m represents the number of sampling points, N represents the number of radars, and Z_nosiy N Let i represent the noisy intermediate frequency signal of the Nth radar, where i represents an integer from 1 to m.
[0060] Solve the sensing matrix using formulas Where Num represents an integer from 0 to 255, expressed as Num = [0, 1, 2, ..., 255]; sin represents the sine trigonometric function, c represents the speed of light, f0 represents the radar frequency; D represents the distance between radars, formula: D = L / (N-1), where L represents the radar aperture; Ang represents the deflection angle, formula: Ang = linspace(Angle min ,Angle max ), where max(·) represents taking the maximum value, linespace(·) represents dividing equally, and Angle min Indicates the minimum deflection angle, Angle max This indicates the maximum deflection angle.
[0061] The following describes the adaptive orthogonal matching algorithm flow. In the following flow:
[0062] r t Let φ represent the residual, t represent the number of iterations, φ represent the empty set, K represent the sparsity, B represent the sensing matrix, and Λ represent the sparseness. t Let λ represent the set of indices (column numbers) for t iterations. t b represents the index (column number) found in the t-th iteration. j Represents the j-th column of matrix B. t Indicates by index Λ tThe set of columns of the selected matrix B, θ t It is a column vector, the symbol ∪ represents the union of sets, <·,·> represents the dot product of vectors, arg(min(.)) represents the minimum value of the expression, and arg(max(.)) represents the maximum value of the expression.
[0063] Step 1: Input the test sample x and the training dictionary A, as well as the sparsity K;
[0064] Step 2: Let the initial residual r0 = x, the initial index set Λ0 = φ, the sensing matrix B0 = φ, and the iteration number t = 1, and initialize the signal initially;
[0065] Step 3: Use formula λ t =arg(max(| <r t-1 ,b j >|)),j=1,2,..,N calculate the index, according to the index λ t search;
[0066] Step 4: Λ t =Λ t-1 ∪λ t B t =B t-1 ∪b λt ;
[0067] Step 5: Use the formula Calculate x = B t θ t The least squares solution;
[0068] Step Six: Use the formula Update residual r t ;
[0069] Step 7: Adaptively adjust the sparsity K using the residual norm difference; if t > 1, then use the formula Calculate the residual of the previous generation, if ||r t-1 -r t ||<||r t || t / 2 Then let t = t-1.
[0070] Step 8: If the iteration condition is met, estimate the sparse representation coefficients of the output signal. Otherwise, return to step three and iterate again. Based on the above steps, the noisy intermediate frequency signal is processed, and after converting the intermediate frequency signal from a time domain signal to a frequency domain signal, the position and angle of the object in space can be calculated according to the established positioning model.
[0071] A method for calculating the position and angle of spatial objects based on the frequency domain signal obtained in the second step.
[0072] After processing multiple intermediate frequency signals using an adaptive orthogonal matching algorithm, multiple peak spectra are obtained, which can be used to estimate the position and angle of multiple detection targets. The frequency f of each sampling point is also obtained. IF The expression is Where X represents the Xth sampling point, n represents the number of sampling points, and F s The sampling frequency is expressed by the formula. Calculate the distance R0 between the target and the antenna.
[0073] like Figure 4 As shown in the diagram, the intersection point of the detection distances of different antenna ports is a schematic diagram. After obtaining the distance, the angular relationship between the antenna and the object is analyzed. Assuming that the distances between the same target detected by antenna one and antenna two and the antenna are R1 and R2 respectively, the intersection point of two circles with radii of R1 and R2 is taken as the target.
[0074] like Figure 5 As shown, based on the above analysis, after calculating the distances R1 and R2 between the target and the two antennas, the included angle is further calculated using the law of cosines, and the relative position between the target and the antennas can then be determined.
[0075] Establish a coordinate system by connecting antenna one and antenna two as the x-axis, and choosing any point along the line as the origin. Let d be the distance from antenna one to the origin. a The distance from antenna 2 to the origin is d. b The distance between antenna one and antenna two is d. 12 The angle between the line connecting antenna one and the target and the x-axis is θ1, expressed as: The angle between the line connecting antenna 2 and the target and the x-axis is θ2, expressed as: The target's ordinate is R1sinθ1 or R2sinθ2, and its abscissa is R1cosθ1-d. a or d a -R2cosθ2.
[0076] The process described above, which obtains the relative distances and angles of multiple objects to radar based on radar intermediate frequency signals, is called super-resolution. An adaptive orthogonal matching algorithm is used to obtain the signal strength of the detected target at each radar's corresponding distance. Figure 6 It can be seen that the signal strength distribution with noise fluctuates quite violently. This violent fluctuation can be clearly seen at a distance of 7. This violent fluctuation will cause great interference in Fourier transform and OMP angle measurement.
[0077] Process the signal using the original OMP, such as Figure 7 As shown, it is clear that the Fourier transform algorithm cannot separate the peaks. The OMP algorithm depends on the sparsity of the signal, resulting in singularities of single or multiple peaks. Its residual norm will stagnate when the signal sparsity is too high. Figure 8This diagram illustrates the relationship between the OMP residual norm and signal sparsity curves. An adaptive constrained orthogonal matching pursuit algorithm is used, and its performance is compared with the original Fourier transform algorithm. Figure 9 In the middle, the two peaks can be clearly separated, and it can be intuitively seen that the intermediate frequency signal comes from the superposition of two signals of different frequencies, indicating that there are two different detection targets.
[0078] like Figure 10 As shown, data containing Gaussian noise was collected in a multi-target moving scene. The data was verified using publicly available millimeter-wave data. There were two objects in the scene, and a total of 86 antenna data were collected. The positions of the two targets were calculated by different antennas. The data was processed by an adaptive constrained orthogonal matching tracking algorithm to obtain the separation spectrum of the objects and calculate the distance and angle information of the targets.
[0079] To verify the super-resolution algorithm localization results, the angle and distance coordinates of the two targets are converted into polar coordinates, as shown below. Figure 11 , Figure 12 As shown. In two-dimensional polar coordinates, the positions of the two targets are (7.32923, -0.02020) and (7.29955, 0.04329), respectively.
[0080] Although the present invention has been disclosed above with reference to preferred embodiments, the embodiments and accompanying drawings are not intended to limit the invention. Any person skilled in the art can make various changes or modifications without departing from the spirit and scope of the invention, and these changes will also be within the protection scope of the invention. Therefore, the protection scope of the present invention should be defined by the scope of the claims of this application.
Claims
1. An adaptive constrained orthogonal matching pursuit algorithm, characterized in that, Includes the following steps: Step 1: After the radar antenna transmits a pulse signal, it performs down-conversion reception at the receiving end to obtain the intermediate frequency signal; Step 2: Process the noisy intermediate frequency signal based on the adaptive orthogonal matching algorithm, and convert the intermediate frequency signal from the time domain signal to the frequency domain signal; Step 2.1: Input test sample And the training dictionary A, and the sparsity K; Step 2.2: Let the initial residual r0 = The initial index set Λ0=φ, the sensing matrix B0=φ, the iteration number t=1, and the signal is initialized. Step 2.3: Use the formula Calculate the index, according to index λ t Search, where N represents the number of radars, and max(·) represents taking the maximum value; Step 2.4: The set of indices for the t-th iteration The sensing matrix in the t-th iteration ; Step 2.5: Use the formula calculate The least squares solution; Step 2.6: Use the formula Update residuals ; Step 2.7: Adaptively adjust the sparsity K using the residual norm difference; if Then use the formula Calculate the residual of the previous generation, if Then let ; Step 2.8: If the iteration condition is met, then estimate the sparse representation coefficients of the output signal. Otherwise, return to step three and iterate again; in, Let represent the residual, t represent the number of iterations, φ represent the empty set, K represent the sparsity, and B represent the sensing matrix. This represents the set of indices for iteration t. This represents the index found in the t-th iteration. Represents the j-th column of matrix B. Indicates by index The set of columns selected for matrix B, It is a column vector, symbol This represents the set union operation. This indicates the expression to calculate the dot product of vectors; arg(min()) indicates the expression to calculate the minimum value; and arg(max()) indicates the expression to calculate the maximum value. Step 3: Based on the frequency domain signal obtained in Step 2, calculate the position of the object in space and its relative angle to the antenna to achieve super-resolution.
2. The adaptive constrained orthogonal matching pursuit algorithm according to claim 1, characterized in that, In step 1: transmitting a signal After the antenna transmits, it will come into contact with a space object and emit a signal, thus obtaining an echo signal. The intermediate frequency signal is obtained by mixing and filtering the two signals, as follows: Assume the transmitted signal The expression is The signal frequency is ,in Indicates the frequency modulation slope, and the range of time t. Duration The FMCW signal on the signal is called a frequency modulation signal; Assuming echo signal The expression is ,in , Indicates echo delay, Indicates the distance to the target. Represents the speed of light. Indicates the radar frequency; within one frequency modulation cycle, the transmitted signal and the echo signal are in time... Since there is overlap, the expression for the output intermediate frequency signal after the mixer is: .
3. The adaptive constrained orthogonal matching pursuit algorithm according to claim 1, characterized in that: When processing noisy intermediate frequency signals, the first step is to select test samples. And construct a training dictionary A, using different radar signal strengths and formulas Selecting test samples Where m represents the number of sampling points and N represents the number of radars. This represents the noisy intermediate frequency signal of the Nth radar. Represents integers from 1 to m.
4. The adaptive constrained orthogonal matching pursuit algorithm according to claim 3, characterized in that: Construct a training dictionary using formulas Where Num represents an integer from 0 to 255, the expression is: sin represents the sine trigonometric function, and c represents the speed of light. D represents the radar frequency; D represents the distance between radars, formula: L represents the radar aperture; Ang is the deflection angle, formula: In this context, linespace (·) represents equal division, and angle... min Indicates the minimum deflection angle, Angle max This indicates the maximum deflection angle.
5. The adaptive constrained orthogonal matching pursuit algorithm according to claim 2, characterized in that, In step 3: The distance R0 between the target and the antenna is expressed by the formula: ,in The expression is , This represents the sampling point corresponding to the peak value in the spectrum, where n represents the number of sampling points. Indicates the sampling frequency.
6. The adaptive constrained orthogonal matching pursuit algorithm according to claim 5, characterized in that: Assume that the distances between the same target detected by antennas 1 and 2 and the antennas are R1 and R2, respectively. Establish a coordinate system with the intersection of two circles with radii R1 and R2 as the target, connecting antennas 1 and 2 as the x-axis, and arbitrarily choosing a point on the line connecting them as the origin. Let the distance from antenna 1 to the origin be d. a The distance from antenna 2 to the origin is d. b The distance between antenna one and antenna two is d. 12 The angle between the line connecting antenna one and the target and the x-axis is... The expression is The angle between the line connecting antenna two and the target and the x-axis is... The expression is The target's vertical coordinate is or The x-axis is or .
7. The adaptive constrained orthogonal matching pursuit algorithm according to claim 2, characterized in that: The millimeter-wave radar, based on 86 antenna elements, collects indoor detection data. The millimeter-wave radar operates at a frequency of 76-81 GHz.
Citation Information
Patent Citations
Improved sparseness self-adaptive matching pursuit algorithm based on dichotomy
CN107612555A
Fast ramp mode LFMCW vehicle-mounted radar signal processing method based on compressed sensing
CN113189576A
