Multi-objective direction finding method based on DBF+k nearest neighbors

Abstract

The application discloses a multi-target direction finding method based on DBF+K neighbors, relates to the field of radio direction finding technology, and solves the problem that the CNN wave arrival angle estimation method can only realize single-target direction finding. The method comprises the following steps: first, orthogonal double-channel sampling is performed on the radiation source signals received by each array element to perform orthogonal conversion and digital sampling processing; second, feature extraction is performed to extract feature parameters for direction finding; third, digital beam forming is performed to realize spatial filtering and angle preliminary classification; fourth, weighted distance calculation is performed to calculate the weighted distance between the signal to be direction found and the training set sample; fifth, sample sorting is performed to sort the sample data according to the distance; sixth, candidate angles are determined, and the K samples with the smallest distance are selected cumulatively to determine the candidate angles of different K values; and seventh, multi-target direction finding based on DBF+K neighbors is performed, and based on the DBF+K neighbor algorithm model, multi-target direction finding is realized simultaneously on the basis of learning and training.

Description

Technical Field

[0001] This invention relates to the field of radio direction finding technology, specifically to a multi-target direction finding method based on DBF+K nearest neighbor. Background Technology

[0002] Direction finding of radiation source targets is a crucial aspect of electronic reconnaissance equipment. It involves using received target signals and processing them to estimate the direction of arrival. Currently, intelligent direction finding methods have become a research hotspot. The patent "Arbitrary Array Angle of Arrival Estimation Method Based on Deep Learning" proposes an array signal direction finding method based on convolutional neural networks (CNNs). This method, for any given multi-element array, extracts feature information such as the phase difference between array elements from the sampled data of each element. Based on deep learning, it achieves rapid and high-precision direction finding of radiation source targets.

[0003] However, the aforementioned CNN direction finding method can only perform direction finding for a single radiation source target. In the case of multiple targets simultaneously, multiple signals overlap and characteristic parameters such as phase difference between array elements will change, thus leading to direction finding errors and difficulties in direction finding. Based on the above shortcomings, this invention proposes a multi-target direction finding method based on DBF+K nearest neighbors. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a multi-target direction finding method based on DBF+K nearest neighbors. This method addresses the problem that CNN angle of arrival estimation methods can only achieve single-target direction finding. It utilizes digital beamforming (DBF) technology to achieve spatial filtering and initial angle classification, and then uses the K nearest neighbor algorithm to achieve high-precision direction finding of multiple targets simultaneously based on learning and training.

[0005] To achieve the above objectives, an embodiment of the first aspect of the present invention proposes a multi-target direction finding method based on DBF+K nearest neighbors, comprising the following steps:

[0006] Step 1: Orthogonal dual-channel sampling: Orthogonal transformation and digital sampling processing are performed on the radiation source signals received by each array element to obtain I and Q dual-channel sampled data of the received signal; the sampled data is used for feature extraction and digital beamforming;

[0007] Step 2: Feature extraction: Extract feature parameters for direction finding, including the relative amplitude and covariance features of each array element;

[0008] Step 3: Digital beamforming and spatial filtering: Based on the simultaneous multi-beam generation of DBF, spatial filtering and initial angle classification are achieved;

[0009] Step 4: Weighted distance calculation, calculate the weighted distance between the signal to be measured and the training set samples;

[0010] Step 5: Sort the sample data according to the distance between them;

[0011] Step 6: Determine candidate angles by iteratively selecting the K samples with the smallest distances to determine candidate angles for different K values;

[0012] Step 7: Multi-target direction finding based on DBF+K nearest neighbors. Based on the DBF+K nearest neighbor algorithm model, simultaneous multi-target direction finding is achieved through learning and training.

[0013] Further, wherein the first step includes:

[0014] Assume the receiving array antenna has L elements, the incident signals have M elements, and the array output is represented by the vector x(t) = [x1(t), x2(t), ..., x...]. L (t)] T The expression is as follows:

[0015]

[0016] Where A = [a1, a2, ... a M ] is the array response matrix corresponding to M incident signals. The incident angle θ of the i-th signal i The corresponding array response vector.

[0017] Furthermore, the second step includes:

[0018] If the signals are independent of each other, then the covariance matrix of the array output is:

[0019] R = E[x(t)x H (t)]

[0020] =AE[s(t)s H (t)]A H +E[v(t)v H (t)]

[0021] =AR s A H +R v

[0022] Where R s and R v These represent the signal covariance matrix and the noise covariance matrix, respectively.

[0023] Furthermore, the third step includes:

[0024] The output of the digital beamformer is a weighted sum of the observation signals of each element; if single-beamforming is performed simultaneously, the weighting vector w for each element is w = [w1 w2 … w L ] T The output of the digital beamformer is:

[0025] y(t)=w H x(t)

[0026] If multiple beams are formed simultaneously, multiple weighted sums are performed and then added together to form multiple different receiving beams.

[0027] Furthermore, the fourth step includes:

[0028] Let the angle classification be P, and the number of sample data for each angle in the training set be N. p If there are 10 samples and the total number of samples is N, then N = N1 + N2 + ... + N P Let Q be the number of signal feature parameters used for direction finding after feature extraction, and let the direction-finding angle y be related to the x-axis of each i-th sample. i The distance is calculated using a weighted distance, which is defined as follows:

[0029]

[0030] Where, x i Let α represent the i-th sample data, and let the superscript j represent the j-th feature parameter. j The weighting coefficients for the j-th feature parameter are α1 + α2 + ... + α. Q =1.

[0031] Furthermore, the sixth step includes:

[0032] Given a current value of K, select the K samples with the smallest distance; assuming that these K samples consist of D angles, then these D angles are the candidate angles for the current value of K.

[0033] Within the allowable range of K value variation, select multiple different K values, repeat the calculation, and determine the candidate angles for different K values ​​one by one.

[0034] Furthermore, the seventh step includes:

[0035] Spatial filtering is performed on the multi-beam formed by DBF, the frequency of candidate angles within the current beam angle is counted, and the angle with the highest frequency is selected as the direction finding result within the current beam angle, thus realizing simultaneous multi-target direction finding based on DBF+K nearest neighbor.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] This invention addresses the limitation of CNN angle-of-arrival estimation methods, which can only achieve single-target direction finding. It employs orthogonal dual-channel sampling to perform orthogonal transformation and digital sampling processing on the radiation source signals received by each array element. The sampled data is used for feature extraction and digital beamforming. Further feature parameters for direction finding are extracted, and digital beamforming is performed to achieve spatial filtering and initial angle classification. Based on DBF+K nearest neighbor multi-target direction finding, and using the DBF+K nearest neighbor algorithm model, high-precision direction finding of multiple targets simultaneously is achieved through learning and training. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a block diagram illustrating the principle of the DBF+K nearest neighbor-based multi-target direction finding method of this invention.

[0040] Figure 2 This is a block diagram illustrating the principle of the K-nearest neighbor algorithm in this invention.

[0041] Figure 3 This is a schematic diagram of the principle of multi-beam digital beamforming in this invention. Detailed Implementation

[0042] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] like Figure 1 As shown, the multi-target direction finding method based on DBF+K nearest neighbors includes the following steps:

[0044] Step 1: Orthogonal dual-channel sampling. The radiation source signals received by each array element undergo orthogonal transformation and digital sampling processing to obtain I and Q channel sampled data of the received signal. This sampled data is used for feature extraction and digital beamforming. Specifically:

[0045] Assume the receiving array antenna has L elements, the incident signals have M elements, and the array output is represented by the vector x(t) = [x1(t), x2(t), ..., x...]. L (t)] T The expression is as follows:

[0046]

[0047] Where A = [a1, a2, ... a M ] is the array response matrix corresponding to M incident signals. The incident angle θ of the i-th signal i The corresponding array response vector;

[0048] Step 2: Feature extraction, extracting feature parameters for direction finding, including the relative amplitude and covariance features of each array element;

[0049] Generally, the signals are independent of each other, so the covariance matrix of the array output is:

[0050] R = E[x(t)x H (t)]

[0051] =AE[s(t)s H (t)]A H +E[v(t)v H (t)]

[0052] =AR s A H +R v

[0053] Where R s and R v These represent the signal covariance matrix and the noise covariance matrix, respectively.

[0054] The relative amplitude of each array element is related to the incident direction of the signal, and the covariance matrix retains the angular information of all targets. Therefore, this invention uses the relative amplitude and covariance characteristics of each array element as characteristic parameters for direction finding.

[0055] Step 3: Digital beamforming and spatial filtering, based on simultaneous multi-beam generation using DBF, to achieve spatial filtering and initial angle classification; specifically:

[0056] The output of the digital beamformer is a weighted sum of the observation signals of each element; simultaneously, in single-beamforming, the weighted vector w = [w1 w2 … w] of each element. L ] T The output of the digital beamformer is:

[0057] y(t)=w H x(t)

[0058] Simultaneous multi-beamforming, by performing multiple weighted sums simultaneously and then adding them together, can create multiple different receiving beams. The principle of simultaneous multi-beam digital beamforming is as follows: Figure 3 As shown;

[0059] Step 4: Weighted distance calculation, calculate the weighted distance between the signal to be measured and the training set samples; the specific calculation steps are as follows:

[0060] Let the angle classification be P, and the number of sample data for each angle in the training set be N. p If there are 10 samples and the total number of samples is N, then N = N1 + N2 + ... + N P (For example, if the direction finding range is 360° and the angular resolution is 0.1°, then the number of angle classifications is 3600. If there are 100 samples per angle, then the total number of samples is 3.6 × 10⁻⁶.) 5 ).

[0061] Let Q be the number of signal feature parameters used for direction finding after feature extraction, and let the direction-finding angle y be related to the x-axis of each i-th sample. i The distance is calculated using a weighted distance, which is defined as follows:

[0062]

[0063] Where, x i Let α represent the i-th sample data, and let the superscript j represent the j-th feature parameter. j The weighting coefficients for the j-th feature parameter are α1 + α2 + ... + α. Q =1;

[0064] Step 5: Sort the sample data according to the distance between them;

[0065] Step 6: Determine candidate angles by iteratively selecting the K samples with the smallest distances to determine candidate angles for different K values; specifically:

[0066] For a given value of K, select the K samples with the smallest distance. Assume that these K samples are composed of D angles (multiple samples may be data samples with the same angle, obviously D≤K), then these D angles are the candidate angles for the current value of K.

[0067] Within the allowable range of K value variation, select multiple different K values, repeat the calculation, and determine the candidate angles for different K values ​​one by one;

[0068] Step 7: Multi-target direction finding based on DBF+K nearest neighbors. Based on the DBF+K nearest neighbor algorithm model, and after learning and training, simultaneous multi-target direction finding is achieved, specifically as follows:

[0069] Spatial filtering is performed on the multi-beam formed by DBF, the frequency of candidate angles within the current beam angle is counted, and the angle with the highest frequency is selected as the direction finding result within the current beam angle, ultimately realizing simultaneous multi-target direction finding based on DBF+K nearest neighbor.

[0070] The method in this invention, based on coarse direction finding using digital beamforming, uses K-nearest neighbors for fine direction finding, achieving intelligent multi-target direction finding; at the same time, the multi-target direction finding accuracy is high, with a direction finding accuracy better than 0.2 degrees;

[0071] To verify the performance of this method, we selected a 9-element uniform linear array with a spacing of half a wavelength and conducted a multi-target direction finding statistical experiment using the method of this invention for multiple radiation source targets with different numbers of radiation sources. The experimental statistical results are shown in Table 1.

[0072] Table 1

[0073]

[0074] Experimental results show that the method of the present invention can estimate the angle of arrival (AHA) of multiple targets simultaneously, and the AHA estimation accuracy is high, with the AHA estimation accuracy of multiple targets being better than 0.2 degrees.

[0075] K-Nearest Neighbors Principle

[0076] The K-Nearest Neighbors (KNN) algorithm is a fundamental method for solving classification and regression problems, and it is one of the most typical analogy learning methods. Analogy learning compares two things, finds their similarities at an abstract level, and uses these similarities to map one thing to another. The KNN algorithm calculates the distance between a sample and the points in the training set, finds the K points with the smallest distance values, and these K points are the sample's K-nearest neighbors. The class to which the sample belongs is determined based on the class of the K-nearest neighbors, with the class that appears most frequently being the class to which the sample belongs.

[0077] The principle of the K-nearest neighbor algorithm is as follows: Figure 2 As shown. Figure 2 In the diagram, there are two different types of sample data, represented by small squares and small triangles, respectively. The data marked by the circle in the center of the diagram is the data to be classified. The category of this unclassified data will vary significantly depending on the value of K.

[0078] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.

[0079] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0080] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-target direction finding method based on DBF+K nearest neighbors, characterized in that, Includes the following steps: Step 1: Orthogonal dual-channel sampling: Orthogonal transformation and digital sampling processing are performed on the radiation source signals received by each array element to obtain I and Q dual-channel sampled data of the received signal; the sampled data is used for feature extraction and digital beamforming; Step 2: Feature extraction: Extract feature parameters for direction finding, including the relative amplitude and covariance features of each array element; Step 3: Digital beamforming and spatial filtering: Based on the simultaneous multi-beam generation of DBF, spatial filtering and initial angle classification are achieved; Step 4: Weighted distance calculation, calculate the weighted distance between the signal to be measured and the training set samples; Step 5: Sort the sample data according to the distance between them; Step 6: Determine candidate angles by iteratively selecting the K samples with the smallest distances to determine candidate angles for different K values; Step 7: Multi-target direction finding based on DBF+K nearest neighbors. Based on the DBF+K nearest neighbor algorithm model, and after learning and training, simultaneous multi-target direction finding is achieved; including: Spatial filtering is performed on the multi-beam formed by DBF, the frequency of candidate angles within the current beam angle is counted, and the angle with the highest frequency is selected as the direction finding result within the current beam angle, thus realizing simultaneous multi-target direction finding based on DBF+K nearest neighbors.

2. The multi-target direction finding method based on DBF+K nearest neighbors according to claim 1, characterized in that, in, The first step includes: Assume the receiving array antenna has L elements, the incident signal has M incident signals, and the array output uses a vector... The expression is as follows: ； in It is the array response matrix corresponding to M incident signals. It is the incident angle of the i-th signal. The corresponding array response vector.

3. The multi-target direction finding method based on DBF+K nearest neighbors according to claim 2, characterized in that, The second step includes: If the signals are independent of each other, then the covariance matrix of the array output is: ； in and These represent the signal covariance matrix and the noise covariance matrix, respectively.

4. The multi-target direction finding method based on DBF+K nearest neighbors according to claim 2, characterized in that, The third step includes: The output of a digital beamformer is a weighted sum of the observation signals from each element; if single-beamforming is performed simultaneously, the weighted vector of each element is... The output of the digital beamformer is: ； If multiple beams are formed simultaneously, multiple weighted sums are performed and then added together to form multiple different receiving beams.

5. The multi-target direction finding method based on DBF+K nearest neighbors according to claim 4, characterized in that, The fourth step includes: Let the angle be classified as The training set contains sample data for each angle. There are [number] samples, and the total number of samples is [number]. ,but Let Q be the number of signal feature parameters used for direction finding after feature extraction. ， Angle to be measured With each i sample The distance is calculated using a weighted distance, which is defined as follows: ； in, Represents the i-th sample data, with superscript. Indicates the first One feature parameter, For the first The weighting coefficients of each feature parameter, .

6. The multi-target direction finding method based on DBF+K nearest neighbors according to claim 5, characterized in that, The sixth step includes: Given a current value of K, select the K samples with the smallest distance; assuming that these K samples consist of D angles, then these D angles are the candidate angles for the current value of K. Within the allowable range of K value variation, select multiple different K values, repeat the calculation, and determine the candidate angles for different K values ​​one by one.

Images