Target coming direction estimation method, system and device based on compressed sensing in single-antenna scene and medium

Abstract

A compressed sensing-based target incoming direction estimation method, system, device and medium in a single-antenna scene, and the method comprises the following steps: obtaining power measurement values of corresponding received signals when a receiving antenna points to different directions of a space, forming a vector p = [p1, p2,..., pL] T by using the obtained power measurement values, and estimating DOA of the received signals; constructing a single-antenna rotation sweeping direction finding model; according to the constructed single-antenna rotation space-sweeping direction-finding model, representing the single-antenna rotation space-sweeping direction-finding model by adopting a compressed sensing method; applying an OMP algorithm to iteratively solve a plurality of steering vectors which are most matched with antenna receiving data, obtaining a grid division scheme of a corresponding index set combined with a space angle, and further obtaining a corresponding DOA estimation result; the system, the device and the medium convert a DOA direction finding problem in a single-antenna scene into a signal reconstruction problem based on a compressed sensing theory based on the method, and DOA estimation is obtained by using an OMP algorithm, so that the calculation complexity is effectively reduced, and meanwhile, the direction finding precision is effectively improved; the problems that in DOA estimation, the complexity of a multi-antenna system is high, and the precision of a traditional single-antenna algorithm is limited are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of DOA estimation, and in particular relates to a method, system, device and medium for estimating a target direction of arrival based on compressed sensing in a single antenna scenario. Background Art

[0002] The application of compressed sensing in Direction of Arrival (DOA) estimation is a new technology developed in recent years. Its purpose is to recover the signal from the received data by using the spatial sparsity of the signal without prior information. Compressed sensing uses the sparsity of the signal to recover the high-dimensional sparse signal in the low-dimensional measurement space through optimization algorithms. Even in the presence of multiple co-frequency signal sources, as long as the signal is sparse in a certain transform domain, compressed sensing can still effectively separate and reconstruct the information of each signal source. In addition, compressed sensing considers the information of all signal sources at the same time through a global optimization method, rather than processing point by point. The use of this global information enables compressed sensing to more effectively identify and separate each signal source when processing co-frequency signals, reducing the impact of local interference on overall reconstruction. In other words, compressed sensing is not affected by the nature of the target, and accurate DOA estimation can be performed regardless of whether the multiple targets are co-frequency. Moreover, compared with various traditional DOA estimation methods, compressed sensing theory can also produce good estimation results under unfavorable conditions such as low signal-to-noise ratio and small number of samples.

[0003] The basic principle of existing direction finding based on compressed sensing is: define the data obtained from the array in the multi-antenna system as the observation signal, and define the array flow matrix corresponding to the subdivided grid as the sensing matrix. Since the signal and the angle interval correspond one to one, the accurate estimation of the azimuth of the incident signal can be completed. Wang Tiedan disclosed an application of compressed sensing technology in array direction finding [University of Electronic Science and Technology of China [2025-01-16].DOI:CNKI:CDMD:1.1013.335696], which can combine compressed sensing theory with array direction finding, and on the basis of compressed sensing theory, make certain improvements to the recovery algorithm for several special models to achieve better recovery effect. However, most of the current direction finding technologies based on compressed sensing focus on the research of multi-antenna systems. However, the implementation of multi-antenna systems often requires complex hardware configuration, strict channel calibration, and is accompanied by high power consumption and cost. These requirements are usually difficult to meet in resource-constrained environments. Summary of the invention

[0004] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a target direction estimation method, system, device and medium based on compressed sensing in a single antenna scenario, construct a compressed sensing framework based on single antenna radiation characteristic modeling, and use the orthogonal matching pursuit (Orthogonal Matching Pursuit, OMP) algorithm to perform DOA estimation for multiple targets; the present invention transforms the DOA direction finding problem in a single antenna scenario into a signal reconstruction problem based on compressed sensing theory, and proposes to use the OMP algorithm to solve it, which effectively reduces the computational complexity and effectively improves the direction finding accuracy, solving the problems of high complexity of multi-antenna systems and limited accuracy of traditional single antenna algorithms in DOA estimation.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] A target direction estimation method based on compressed sensing in a single antenna scenario includes the following steps:

[0007] Step 1: Obtain the power measurement values ​​of the received signal when the receiving antenna points to different directions in space, and use the obtained power measurement values ​​to form a vector p = [p 1 ,p 2 ,..,p L ] T , estimate the DOA of the received signal; construct a single antenna rotating sweep direction finding model;

[0008] Step 2, combining the single-antenna rotating sweeping direction finding model constructed in step 1, and using a compressed sensing method to represent the single-antenna rotating sweeping direction finding model;

[0009] Step 3: Based on step 2, the OMP algorithm is used to iteratively solve multiple steering vectors that best match the antenna reception data, obtain the grid division scheme of the corresponding index set combined with the spatial angle, and then obtain the corresponding DOA estimation result.

[0010] The specific method of step 1 is:

[0011] 1.1 When the antenna points to different directions, the power of the received signal is measured by a single antenna receiver capable of measuring the power of the received signal. A set of power values ​​measured by the receiver when the antenna points to different directions in space is taken to form a received power vector; the received signal is modeled as the sum of all signal sources affected by the gain of the receiving antenna in different directions, expressed as:

[0012]

[0013] In the receiving signal model, To point the antenna For time from θ k The signal gain incident on the antenna, s k(t) is the kth incident signal, η(t) is the receiver noise, and K is the number of signal sources. When the spatial pointing direction of the antenna changes continuously, its gain to the incoming signal will also change accordingly. l (t) represents the spatial orientation of the antenna. The received signal at:

[0014]

[0015] Considering that the K signal sources in a finite short period T have no information change and are unrelated to each other, the power p of the signal received by the receiver in a period T is l Approximately:

[0016]

[0017] in, is regarded as the spatial power gain of the antenna to the incoming signal, and it depends on both the direction of the receiving antenna and the incident direction of the incoming signal; γ k is the power of the signal source reaching the receiving antenna, n l is the receiver noise power:

[0018]

[0019] The power measurement values ​​of the received signal corresponding to the receiving antenna pointing to different directions in space are obtained, and the obtained power measurement values ​​are used to form a vector p = [p 1 ,p 2 ,..,p L ] T , estimate the DOA of the received signal, that is, θ=[θ 1 ,θ 2 ,..,θ K ] T , where L is the number of received power observation samples of the direction finding system, and K is the number of incoming signal sources;

[0020] 1.2 Describe the antenna characteristics and power estimation model:

[0021] The directivity of the direction-finding receiving antenna indicates the receiving antenna's ability to receive electromagnetic wave signals from different directions. The directivity a(θ) of the direction-finding receiving antenna is fitted and expressed.

[0022]

[0023] Among them, m is called the mode of the receiving antenna, a m is called the size of the pattern, M represents the antenna accuracy; if and only if When a(θ) is a real function, and a(θ) satisfies:

[0024]

[0025] In the actual signal processing process, the power estimation of the received signal is achieved by measuring the average power value of the discrete samples of the power of L received signals. Therefore, for the power p of the received signal l Rewrite as follows:

[0026]

[0027] So far, the single-antenna rotating scanning direction finding model has been established, that is

[0028]

[0029] The specific method of step 2 is as follows:

[0030] Suppose there are K far-field narrowband signals incident on a single antenna. Then the signal model received after the antenna rotates one week is that in the range of 0° to 360° for the overall airspace, the airspace range is equally divided into N grid cells, and N azimuth angles {θ 1 , θ 2 ,..., θ N} are corresponding. Each θ n (n = 1, 2,..., N) will correspond to a potential incident signal s n . Define a symbol to represent that there is a real target in the direction of the incident signal, and the corresponding s n is the complex amplitude, a non-zero value. Define another symbol to represent that there is no target in the direction of the incident signal, and the corresponding s n is 0; when the adjacent angle interval Δθ = 360° / N in the airspace range is small enough, the entire airspace will be divided into many grid cells, and the number of grid cells is much larger than the number of incident signals (N >> K). Therefore, the airspace signal is regarded as an N×1-dimensional sparse signal with a sparsity of K;

[0031] Denote the angle pointed by the antenna as the observation angle θ n (1 < n < N). Combining with the single-antenna rotating scanning direction finding model in step 1, we have:

[0032]

[0033] The vector form of equation (10) is:

[0034] p = Ay + n (11)

[0035] where p ∈ C L×1 represents the received signal power vector; A ∈ C L×N refers to the gain of the receiving antenna for L measurement angles and N candidate directions. Among them, each L×1-dimensional steering vector a j(1≤j≤N) corresponds to a target information, and a set of bases formed by the steering vector is used to represent the incoming signal, which is regarded as the sensing matrix in the compressed sensing theory; y∈C N×1 Refers to a sparse signal with sparsity K, in which only K elements are non-zero, corresponding to a limited number of DOA directions, and NK elements are 0; n∈C L×1 is the noise vector;

[0036] The high-dimensional signal y is reconstructed using the low-dimensional power measurement value p and the sensor matrix A, as shown in the following formula:

[0037] min||y|| 0 , st||p-Ay|| 2 ≤ε (12)

[0038] Or: min||y|| 1 , st||p-Ay|| 2 ≤ε (13)

[0039] Among them, ε is a parameter related to noise. By solving the above equations (12) and (13), the positions corresponding to the K non-zero elements in the high-dimensional signal y are obtained.

[0040] The specific method of step 3 is:

[0041] 3.1 Set the initial value of the iteration; Set the initial value of the sparse signal is 0, the compressed signal residual r 0 Equal to the single antenna received power p, index set Λ 0 is an empty set, and the number of iterations k is 1, that is,

[0042]

[0043] 3.2 Extracting dictionary atoms a (column vector) and compressed signal residual r from the sensor matrix A 0 By comparison, the dictionary atom with the highest matching degree is obtained and recorded as f k , that is, to find the steering vector that best matches the antenna received data:

[0044]

[0045] 3.3 Atomic index set Λ 0 Update, search and compress the signal residual r 0 The most relevant dictionary atom index is then imported into the atom index set, i.e., the steering vector that best matches the direction of arrival information is found;

[0046] Λ K =Λ K-1 ∪{f k} (16)

[0047] 3.4 Calculate the sparse coefficients using least squares;

[0048]

[0049] 3.5 Calculate the new residual;

[0050]

[0051] 3.6 Determine whether the iteration process is terminated, that is, determine whether the number of iterations k is less than the number of information sources K. If the number of iterations k is less than the number of information sources K, set k=k+1 and go to step 3.2 to continue iterating; when k=K, end the iteration;

[0052] 3.7 Save and output the estimated results according to The positions of non-zero elements in the grid and the spatial angles are converted into the incident angle of the target using the atomic index set.

[0053] A system for a target direction estimation method based on compressed sensing in a single antenna scenario, comprising:

[0054] The single antenna rotation sweeping model establishment module is used in step 1 to establish the DOA estimation model in the single antenna scenario by constructing the received signal model and antenna characteristics;

[0055] The DOA direction finding module based on compressed sensing in the single antenna scenario is used in step 2. By utilizing the spatial sparsity of the DOA signal in the single antenna rotation sweeping model, the sparse reconstruction model is used to solve the problem under the framework of compressed sensing, thereby converting the direction finding problem into the solution of the signal sparse reconstruction problem under the compressed sensing theory.

[0056] The problem solving module based on the OMP algorithm is used in step 3 to solve the single antenna direction finding problem through the OMP algorithm to achieve the DOA estimation of the target.

[0057] An electronic device for estimating a target direction based on compressed sensing in a single antenna scenario, comprising:

[0058] Memory for storing computer programs;

[0059] A processor is used to implement the method for target estimation based on compressed sensing in a single antenna scenario described in steps 1 to 3 when executing the computer program.

[0060] A computer-readable storage medium stores a computer program, which, when executed by a processor, can perform DOA estimation according to the method for estimating the target based on compressed sensing in a single antenna scenario described in steps 1 to 3.

[0061] Compared with the prior art, the present invention has the following advantages:

[0062] The present invention provides a new technical solution for direction finding based on compressed sensing in a single-antenna scenario: the present invention models the radiation characteristics of the antenna as the sum of a finite number of weighted complex exponentials, and applies the model to the modeling of the received signal power, thereby extracting the component containing DOA information from the power model. The idea of ​​compressed sensing is adopted to convert the power model into a compressed sensing model, and the OMP algorithm is used to estimate the DOA of multiple targets, thereby avoiding the problems of complex hardware configuration, strict channel calibration requirements, high power consumption and high cost brought about by the use of a multi-antenna system. At the same time, it overcomes the problem that the traditional single-antenna system cannot be applied under complex conditions such as low signal-to-noise ratio, limited number of samples and co-frequency of the target source.

[0063] The present invention has higher direction finding accuracy and lower computational complexity: the present invention utilizes the antenna characteristics of the rotating antenna and the power measurement samples, innovatively proposes to use compressed sensing to solve the DOA estimation problem in a single antenna scenario, converts the power model into a compressed sensing model, and uses the OMP algorithm to perform DOA estimation for multiple targets. Compared with other algorithms, the computational complexity is effectively reduced, and the direction finding accuracy is better than other traditional algorithms.

[0064] In summary, the present invention models the antenna radiation characteristics as the sum of finite-term weighted complex exponentials, constructs a compressed sensing framework and uses the orthogonal matching pursuit (OMP) algorithm for DOA estimation, which effectively reduces the hardware cost and computational complexity while improving the direction finding accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is an estimation flow chart based on the goal of compressed sensing in the present invention.

[0066] Figure 2 It is a schematic diagram of the spatial domain grid division of the present invention.

[0067] Figure 3 It is the curve of the direction finding performance of various algorithms simulated by the present invention changing with the increase of signal-to-noise ratio.

[0068] Figure 4 It is the curve of the direction finding performance of various algorithms simulated by the present invention as the number of samples increases.

[0069] Figure 5 It is a curve showing how the direction finding performance of various algorithms simulated by the present invention changes with the increase of the incoming wave direction interval.

[0070] Figure 6 It is the curve of the direction finding performance of various algorithms simulated by the present invention changing with the increase of the number of signal sources. DETAILED DESCRIPTION

[0071] The present invention is further described in detail below in conjunction with the accompanying drawings and simulation experiments.

[0072] See also Figure 1 , a target direction estimation method based on compressed sensing in a single antenna scenario, specifically comprising the following steps:

[0073] Step 1: Obtain the power measurement values ​​of the received signal when the receiving antenna points to different directions in space, and use the obtained power measurement values ​​to form a vector p = [p 1 ,p 2 ,..,p L ] T , estimate the DOA of the received signal; the specific method of constructing a single antenna rotating sweep direction finding model is:

[0074] 1.1 When the antenna points to different directions, the power of the received signal is measured by a single antenna receiver capable of measuring the power of the received signal. A set of power values ​​measured by the receiver when the antenna points to different directions in space is taken to form a received power vector; the received signal is modeled as the sum of all signal sources affected by the gain of the receiving antenna in different directions, expressed as:

[0075]

[0076] In the receiving signal model, To point the antenna For time from θ k The signal gain incident on the antenna, s k (t) is the kth incident signal, η(t) is the receiver noise, and K is the number of signal sources. When the spatial pointing direction of the antenna changes continuously, its gain to the incoming signal will also change accordingly. l (t) represents the spatial orientation of the antenna. The received signal at:

[0077]

[0078] Considering that the K signal sources in a finite short period T have no information change and are unrelated to each other, the power p of the signal received by the receiver in a period T is l Approximately:

[0079]

[0080] in, is regarded as the spatial power gain of the antenna to the incoming signal, and it depends on both the direction of the receiving antenna and the incident direction of the incoming signal; γ k is the power of the signal source reaching the receiving antenna, n l is the receiver noise power:

[0081]

[0082] Thus, the power measurement values ​​of the received signal corresponding to the receiving antenna pointing to different directions in space are obtained, and the group of power observation values ​​is formed into a vector p = [p 1 ,p 2 ,..,p L ] T , the purpose is to estimate the DOA of the received signal, that is, θ = [θ 1 ,θ 2 ,..,θ K ] T , where L is the number of received power observation samples of the direction finding system, and K is the number of incoming signal sources;

[0083] 1.2 Describe the antenna characteristics and power estimation model:

[0084] The directivity of the direction-finding receiving antenna indicates the receiving antenna's ability to receive electromagnetic wave signals from different directions. The directivity a(θ) of the receiving antenna is fitted to represent the receiving antenna.

[0085]

[0086] Among them, m is called the mode of the receiving antenna, a m It is called the size of the pattern. Generally, M is 5 to 10, which can represent the antenna with high accuracy. When a(θ) is a real function, and a(θ) satisfies:

[0087]

[0088] In the actual signal processing process, in order to reduce the complexity and cost of the direction finding system construction, the power estimation of the received signal is achieved by measuring the average power value of L discrete samples of the power of the received signal. l Rewrite as follows:

[0089]

[0090] At this point, the single antenna rotation sweep direction finding model is established, that is,

[0091]

[0092] Step 2: Combined with the single-antenna rotating sweeping direction finding model constructed in step 1, a signal sparse reconstruction model based on compressed sensing theory is used to represent the single-antenna rotating sweeping direction finding model; the specific method is:

[0093] Suppose K far-field narrowband signals are incident on a single antenna. The signal model received after the antenna rotates one circle is as follows: Figure 2, within the range of 0° to 360° facing the entire airspace, the airspace is equally divided into N grids at equal intervals, and corresponds to N azimuth angles {θ 1 , θ 2 ,..., θ N}, each θ n (n = 1, 2,..., N) will correspond to a potential incident signal s n , where "·" indicates that there is a real target in the direction of the incident signal, and the corresponding s n is the complex amplitude, a non-zero value, while indicates that there is no target in the direction of the incident signal, and the corresponding s n is 0; when the adjacent angle interval Δθ = 360° / N in the airspace is small enough, the entire airspace will be divided into many grids, and the number of grids is much larger than the number of incident signals (N >> K). Therefore, the airspace signal is regarded as an N×1-dimensional sparse signal with a sparsity of K;

[0094] Denote the angle pointed by the antenna as the observation angle θ n (1 < n < N). Combining with the single-antenna rotation scanning direction-finding model in step 1, we have

[0095]

[0096] The vector form of the above formula is the following formula:

[0097] p = Ay + n

[0098] where, p ∈ C L×1 represents the received signal power vector; A ∈ C L×N refers to the gain of the receiving antenna for L measurement angles and N candidate directions. Each L×1-dimensional steering vector a j (1 ≤ j ≤ N) corresponds to a target information. A set of bases formed by the steering vectors are used to represent the incoming wave signal and are regarded as the sensing matrix in the compressed sensing theory; y ∈ C N×1 refers to a sparse signal with a sparsity of K, where only K elements are non-zero, corresponding to a finite number of DOA directions, and N - K elements are 0; n ∈ C L×1 is the noise vector;

[0099] Use the low-dimensional measurement value p and matrix A to reconstruct the high-dimensional signal y, as shown in the following formula,

[0100] min ||y|| 0 , s.t. ||p - Ay|| 2 ≤ ε

[0101] Or: min ||y|| 1 , s.t. ||p - Ay||2 ≤ε

[0102] Among them, ε is a parameter related to noise. By solving the above equation, the positions corresponding to the K non-zero elements in y are obtained, and then the corresponding DOA estimation results are obtained based on the positions and the grid division scheme.

[0103] Step 3: Based on step 2, the OMP algorithm is used to iteratively solve multiple steering vectors that best match the antenna reception data, obtain the grid division scheme of the corresponding index set combined with the spatial angle, and then obtain the corresponding DOA estimation result. The specific method is:

[0104] Select the OMP algorithm in the greedy algorithm for iterative reconstruction solution:

[0105] 3.1 Set the initial value of the iteration; Set the initial value of the sparse signal is 0, the compressed signal residual r 0 Equal to single antenna receiving data p, index set Λ 0 is an empty set, and the number of iterations k is 1, that is,

[0106]

[0107] 3.2 Extracting dictionary atoms a (column vector) and compressed signal residual r from the sensor matrix A 0 By comparison, the dictionary atom with the highest matching degree is obtained and recorded as f k , that is, to find the steering vector that best matches the antenna received data:

[0108]

[0109] 3.3 Atomic index set Λ 0 Update, search and compress the signal residual r 0 The most relevant dictionary atom index is then imported into the atom index set, i.e., the steering vector that best matches the direction of arrival information is found;

[0110] Λ K =Λ K-1 ∪{f k}

[0111] 3.4 Calculate the sparse coefficients using least squares;

[0112]

[0113] 3.5 Calculate the new residual;

[0114]

[0115] 3.6 Determine whether the iteration process is terminated. That is, determine whether the number of iterations k is less than the number of information sources K. If it is less than K, set k = k + 1 and repeat process 3.2; when k = K, the iteration ends.

[0116] 3.7 Save and output the estimated results according to The grid partitioning scheme of the position and spatial angle of non-zero elements in the image is used to convert the incident angle of the target using the atomic index set.

[0117] 1. Simulation conditions

[0118] In order to simulate the received signal, the following constraints were imposed on the preconditions in the simulation: the spatial pointing of all experimental antennas varied uniformly between 0° and 359°, the angle domain from 0° to 359° was divided into 360 grids, that is, the angle grid N = 360, and the antenna beam width M = 9. It was assumed that each signal source was independent of each other, and the signal-to-noise ratio relative to the noise signal was 0, that is, the signal amplitude was 1, and all experiments were conducted under Gaussian white noise scenarios.

[0119] In order to illustrate the experimental results, the estimation performance curve of the traditional single-antenna direction finding method based on power measurement under the same conditions is also given for comparison, and the root mean square error (RMSE) of the angle estimation of 1000 Monte Carlo experiments is used as the measurement standard.

[0120]

[0121] Where K is the signal sparsity, Indicates θ in the nth experiment K The estimated value of .

[0122] In order to verify the performance of the proposed method, the estimation performance of the OMP algorithm based on compressed sensing is investigated and analyzed from the aspects of signal-to-noise ratio, estimation time, number of power samples, incoming wave direction interval, number of incoming wave signals, etc. The following comparison algorithms are selected for direction finding performance comparison simulation: the classic annihilation filter method based on the reconstruction problem of sampling signals with finite information rate; the correlation peak DOA estimation method based on directional pattern matching; the OMP algorithm;

[0123] 2. Simulation content and results

[0124] Simulation 1: Assume that two incoming wave signals come from θ 1 =50°,θ 2 =150° direction to the receiving antenna, the number of power samples L = 180, the signal-to-noise ratio of the signal changes from -10dB to 10dB in 2dB steps, the simulation diagram is as follows Figure 3 shown.

[0125] Depend on Figure 3It can be seen that the root mean square errors of the three DOA estimation methods show significant differences with the change of signal-to-noise ratio. The error of OMP estimation is about 10 degrees under low signal-to-noise ratio conditions, showing strong noise resistance. After the signal-to-noise ratio exceeds 0dB, the error is close to 0 degrees, indicating that OMP can achieve high-precision DOA estimation under high signal-to-noise ratio. In contrast, the error of the annihilation filter estimation is extremely large under low signal-to-noise ratio conditions. For example, the RMSE is as high as 90 degrees at -10dB, and even at 10dB, the error can only be reduced to about 10 degrees, showing poor noise resistance and convergence performance. The error of the correlation peak estimation is about 30 degrees at low signal-to-noise ratio, which is better than the annihilation filter estimation, but not as good as the OMP algorithm. However, as the signal-to-noise ratio increases, the RMSE decreases rapidly, approaching the performance of OMP. In general, OMP estimation shows the best noise resistance and estimation accuracy in the entire signal-to-noise ratio range, especially under low signal-to-noise ratio conditions; correlation peak estimation has high accuracy under high signal-to-noise ratio, but its performance is lacking under low signal-to-noise ratio; annihilation filter estimation performs the worst, is more sensitive to noise, and has poor overall performance, especially under low signal-to-noise ratio conditions, the error is significantly higher than the other two methods.

[0126] Simulation 2: Assume that two incoming wave signals come from θ 1 =50°,θ 2 =150° direction reaches the receiving antenna, the number of power samples remains unchanged at L=180, and the signal-to-noise ratio SNR=-5dB. The average duration and angle estimation results of each algorithm DOA estimation are compared under 1000 Monte Carlo simulations. The simulation results are shown in Table 1.

[0127] Table 1.1Table 1DOA estimation algorithm duration estimation table =

[0128]

[0129] According to the table data, it can be seen that there are significant differences in the angle estimation results and average estimation time of the three DOA estimation algorithms under the condition of SNR = -5dB. The RMSE of OMP estimation is 0.6783°, with a small error, showing high estimation accuracy at a low calculation time. The DOA accuracy of correlation peak estimation is close to that of OMP estimation, but the calculation time is hundreds of times that of other algorithms. The RMSE of annihilation filter estimation far exceeds that of other algorithms, and the error is extremely large. In terms of comprehensive accuracy and time, OMP estimation has a much better calculation efficiency than correlation peak estimation while maintaining a high estimation accuracy, showing obvious comprehensive performance advantages.

[0130] Simulation 3: The signal-to-noise ratio remains unchanged at SNR = -5dB. The number of received power samples of the direction finding system increases from 20 to 180 in steps of 20. The two incoming wave signals are respectively from θ 1 =50°,θ 2 =150° direction reaches the receiving antenna, the simulation diagram is as follows Figure 4 shown.

[0131] As can be seen from the figure, as the number of power increases, the estimation of the OMP algorithm is getting closer and closer to the actual DOA, and when the number of power observations is greater than or equal to 100, the DOA estimation value is basically consistent with the actual DOA value, indicating that in this low signal-to-noise ratio measurement environment, the two incoming signal sources can obtain a relatively ideal DOA estimation result by taking more than 100 received power values. The overall trend of the correlation peak estimation method is basically consistent with the OMP algorithm, but the estimation accuracy is slightly inferior to the OMP algorithm, and when the number of samples L increases, the time required for estimation will increase significantly. The root mean square error of the classic annihilation filter method based on the FRI sampling signal reconstruction problem is significantly lower than that of the OMP algorithm in this environment, indicating that this method cannot achieve the expected estimation effect by simply increasing the number of measured received powers, because at this time the restriction of noise on DOA estimation is far greater than the improvement that can be brought about by increasing the number of power samples. Therefore, in a noisy environment, especially under low signal-to-noise ratio conditions, the impact of noise on DOA estimation is the strongest. In summary, the use of the OMP estimation algorithm based on compressed sensing still has a good DOA estimation effect under low signal-to-noise ratio, low time consumption and a small number of power samples.

[0132] Simulation 4: When the number of received power samples L = 180 and the signal-to-noise ratio parameter SNR = -5dB, the direction of one of the incoming signals is fixed to θ 1 =50°, and the direction of the other signal source arriving at the receiving antenna changes at 5° intervals, the DOA estimation effects of the three methods change with the interval between the arrival angles of the two signal sources. The simulation diagram is as follows Figure 5 shown.

[0133] According to simulation Figure 5 It can be seen that for the annihilation filter estimation method, when the arrival angles of the two signal sources are greater than or equal to 30°, there is still a significant deviation between the estimation result and the actual DOA, and this deviation is mainly affected by noise. At the same time, when the signal source arrival angle interval is less than 25°, the estimation error of the correlation peak estimation method remains almost unchanged and the error is extremely large; when the signal source interval reaches 25°, the estimation error is significantly reduced. This shows that when the signal source arrival angle interval is less than 25°, the algorithm has insufficient resolution and it is difficult to effectively distinguish the directions of adjacent signal sources. The OMP estimation algorithm based on compressed sensing gradually reduces the estimation error as the signal source interval increases, and when the arrival angles of the two signal sources are greater than or equal to 30°, the estimation result is almost consistent with the actual DOA. Comprehensive analysis shows that the OMP algorithm exhibits higher angular resolution under the same experimental conditions and can more accurately estimate the direction of the signal source.

[0134] Simulation 5: When the number of received powers remains unchanged at L=180 and the signal-to-noise ratio changes from -10dB to 10dB, the estimated effects of one, two, three, and four incoming signals are statistically analyzed. The signal arrival direction combinations in these three cases are {50°}, {50°, 80°}, {50°, 80°, 130°}, and {50°, 80°, 130°, 150°}, respectively. The simulation diagram is shown in the figure below. Figure 6 As shown in the simulation Figure 6 It can be seen that under the same relative noise intensity, the fewer the number of incoming wave signal sources, the more accurate the DOA estimation result. Under low signal-to-noise ratio conditions, the more the number of signal sources, the greater the deviation of the DOA estimation result from the actual DOA. This is mainly because when the number of incoming wave signal sources increases, interference between signals is introduced, and the coherence between signals may be high (for example, the angles of two signal sources are close). Especially in the case of low signal-to-noise ratio, noise interference aggravates the confusion between signals, resulting in a decrease in the estimation accuracy of the OMP algorithm.

Claims

1. A method for estimating target direction based on compressed sensing in a single antenna scenario, characterized in that: The following steps are involved: Step 1: Obtain the power measurement values ​​of the received signal corresponding to the receiving antenna pointing to different directions in space, and form a vector p = [p1, p2, .., p L ] T , estimate the DOA of the received signal; construct a single antenna rotating sweep direction finding model; Step 2, combining the single-antenna rotating sweeping direction finding model constructed in step 1, and using a compressed sensing method to represent the single-antenna rotating sweeping direction finding model; Step 3: Based on step 2, the orthogonal matching pursuit (OMP) algorithm is used to iteratively solve the steering vector that best matches the antenna reception data, and the target direction is obtained, that is, the corresponding DOA estimation result.

2. According to claim 1, the target direction estimation method based on compressed sensing in a single antenna scenario is characterized in that: The specific method of step 1 is: 1.1 When the antenna points to different directions, the power of the received signal is measured by a single antenna receiver capable of measuring the power of the received signal. A set of power values ​​measured by the receiver when the antenna points to different directions in space is taken to form a received power vector; the received signal is modeled as the sum of all signal sources affected by the gain of the receiving antenna in different directions, expressed as: In the receiving signal model, To point the antenna For time from θ k The signal gain incident on the antenna, s k (t) is the kth incident signal, η(t) is the receiver noise, and K is the number of signal sources. When the spatial pointing direction of the antenna changes continuously, its gain to the incoming signal will also change accordingly. l (t) represents the spatial orientation of the antenna. The received signal at: Considering that the K signal sources in a finite short period T have no information change and are unrelated to each other, the power p of the signal received by the receiver in a period T is l Approximately: in, is regarded as the spatial power gain of the antenna to the incoming signal, and it depends on both the direction of the receiving antenna and the incident direction of the incoming signal; γ k is the power of the signal source reaching the receiving antenna, n l is the receiver noise power: The power measurement values ​​of the received signal corresponding to the receiving antenna pointing to different directions in space are obtained, and the obtained power measurement values ​​are used to form a vector p = [p1, p2, .., p L ] T , estimate the DOA of the received signal, that is, θ=[θ1,θ2,..,θ K ] T , where L is the number of received power observation samples of the direction finding system, and K is the number of incoming signal sources; 1.2 Describe the antenna characteristics and power estimation model: The directivity of the direction-finding receiving antenna indicates the receiving antenna's ability to receive electromagnetic wave signals from different directions. The directivity a(θ) of the direction-finding receiving antenna is fitted and expressed. Among them, m is called the mode of the receiving antenna, a m is called the size of the pattern, M represents the antenna accuracy; if and only if When a(θ) is a real function, and a(θ) satisfies: In the actual signal processing process, the power estimation of the received signal is achieved by measuring the average power value of L discrete samples of the power of the received signal. l Rewrite as follows: At this point, the single antenna rotation sweep direction finding model is established, that is, 3. According to the method for estimating the target direction based on compressed sensing in a single antenna scenario in claim 1, it is characterized in that: The specific method of step 2 is: Suppose K far-field narrowband signals are incident on a single antenna. The signal model received by the antenna after one rotation is within the range of 0° to 360° facing the entire airspace. The airspace is divided into N grids at equal intervals, corresponding to N azimuth angles {θ1,θ2,...,θ N }, each θ n (n=1,2,...,N) will correspond to a potential incident signal s n , a custom symbol is used to indicate that there is a real target in the direction of the incident signal, and the corresponding s n is the complex amplitude, non-zero value, and another custom symbol indicates that there is no target in the direction of the incident signal, and the corresponding s n is 0; when the adjacent angle interval Δθ=360° / N in the spatial domain is small enough, the entire spatial domain will be divided into many grids, and the number of grids is much larger than the number of incident signals (N>>K), so the spatial domain signal is regarded as a sparse signal of N×1 dimension with a sparsity of K; Record antenna The angle of the direction pointing is the observation angle θ n (1 < n < N). Combining with the single-antenna rotation and sweeping direction-finding model in Step 1, we have: The vector form of formula (10) is: p=Ay+n (11) Where p∈C L×1 Represents the received signal power vector; A∈C L×N Refers to the gain of the receiving antenna for L measurement angles and N candidate directions, where each L×1-dimensional steering vector a j (1≤j≤N) corresponds to a target information, and a set of bases formed by the steering vector is used to represent the incoming signal, which is regarded as the sensing matrix in the compressed sensing theory; y∈C N×1 Refers to a sparse signal with sparsity K, in which only K elements are non-zero, corresponding to a limited number of DOA directions, and NK elements are 0; n∈C L×1 is the noise vector; The high-dimensional signal y is reconstructed using the low-dimensional power measurement value p and the sensor matrix A, as shown in the following formula: min||y||0,st||p-Ay||2≤ε (12) Or: min||y||1,st||p-Ay||2≤ε (13) Where ε is a parameter related to noise. By solving the above equation (12) or (13), the positions corresponding to the K non-zero elements in the high-dimensional signal y are obtained.

4. According to claim 1, the target direction estimation method based on compressed sensing in a single antenna scenario is characterized in that: The specific method of step 3 is: 3.1 Set the initial value of the iteration; Set the initial value of the sparse signal is 0, the compressed signal residual r0 is equal to the single antenna received power p, the index set Λ0 is an empty set, and the number of iterations k is 1, that is, 3.2 Extract dictionary atom a (column vector) from sensor matrix A and compare it with compressed signal residual r0 to obtain the dictionary atom with the highest correlation matching degree, denoted as f k , that is, to find the steering vector that best matches the antenna received data: 3.3 Update the atomic index set Λ0, search for the dictionary atomic index that is most relevant to the compressed signal residue r0, and then import it into the atomic index set, that is, find the steering vector that best matches the direction of arrival information; L K =L K-1 ∪{f k } (16) 3.4 Calculate the sparse coefficients using least squares; 3.5 Calculate the new residual; 3.6 Determine whether the iteration process is terminated, that is, determine whether the number of iterations k is less than the number of information sources K. If the number of iterations k is less than the number of information sources K, set k=k+1 and go to step 3.2 to continue iterating; when k=K, end the iteration; 3.7 Save and output the estimated results according to The positions of non-zero elements in the grid and the spatial angles are converted into the incident angle of the target using the atomic index set.

5. A system for target direction estimation based on compressed sensing in a single antenna scenario, characterized in that: include: The single antenna rotation sweeping model establishment module is used in step 1 to establish the DOA estimation model in the single antenna scenario by constructing the received signal model and antenna characteristics; The DOA direction finding module based on compressed sensing in the single antenna scenario is used in step 2. By utilizing the spatial sparsity of the DOA signal in the single antenna rotation sweeping model, the sparse reconstruction model is used to solve the problem under the framework of compressed sensing, thereby converting the direction finding problem into the solution of the signal sparse reconstruction problem under the compressed sensing theory. The problem solving module based on the OMP algorithm is used in step 3 to solve the single antenna direction finding problem through the OMP algorithm to achieve the DOA estimation of the target.

6. An electronic device for target direction estimation based on compressed sensing in a single antenna scenario, characterized in that: include: Memory for storing computer programs; A processor is used to implement the method for target estimation based on compressed sensing in a single antenna scenario described in steps 1 to 3 when executing the computer program.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it is capable of performing DOA estimation according to the method for target direction estimation based on compressed sensing in a single antenna scenario according to claims 1 to 5.

Images