Radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation

Abstract

This invention discloses a radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation. First, an adaptive sparse sensing algorithm is used to process radar pulse descriptor word (PDW) data, transforming the radar pulse signal into a weighted undirected graph. Then, a GraphSAGE network model with a weighted aggregation mechanism is introduced to complete the graph node classification, thereby achieving radar signal sorting. Simulation results show that, when the reconnaissance receiver faces a scenario with high density and uneven distribution of radiation source pulses in a complex electromagnetic environment, this method exhibits higher sorting accuracy and stronger robustness compared to existing GCN-based radar signal sorting algorithms.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, specifically relating to a radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation. Background Technology

[0002] Radar signal sorting refers to the process of classifying interleaved and overlapping radar pulse signals received by a radar receiver into multiple independent radiation source categories. As a crucial step in radar electronic reconnaissance, accurate separation of interleaved and overlapping radar signals is essential for subsequent tasks such as radar signal feature extraction, identification, and threat assessment. With the emergence of new multi-functional radar systems, radars switch between different operating modes and dynamically adjust parameter ranges and modulation types based on actual conditions. This causes traditional signal sorting algorithms to easily classify different operating modes of a single multi-functional radar as multiple radars, resulting in an "over-batch" phenomenon, posing a significant challenge to radar signal sorting. Existing radar signal sorting methods based on Graph Convolutional Networks (GCNs) and Residual Networks (ResNets) can address the "over-batch" problem and performance degradation under small sample constraints to some extent. However, these methods do not consider the global distribution characteristics of radar pulse data during the graph structure construction stage, leading to redundant connections in high-density areas and insufficient topological information in low-density areas. Furthermore, GCNs have limited processing capabilities for large-scale graph structures, resulting in limitations in model generalization and efficiency. When the reconnaissance receiver faces a scenario with high density and uneven distribution of radiation source pulses in a complex electromagnetic environment, the performance of this sorting algorithm drops sharply, indicating certain limitations. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides a radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation. First, an adaptive sparse sensing algorithm is used to process radar pulse descriptor word (PDW) data, transforming the radar pulse signal into a weighted undirected graph. Then, a GraphSAGE network model with a weighted aggregation mechanism is introduced to complete the graph node classification, thereby achieving radar signal sorting. Simulation results show that, when the reconnaissance receiver faces a scenario with high density and uneven distribution of radiation source pulses in a complex electromagnetic environment, this invention's method exhibits higher sorting accuracy and stronger robustness compared to existing GCN-based radar signal sorting algorithms.

[0004] The technical solution adopted by this invention to solve its technical problem is as follows: Step 1: Construct graph structure data using the adaptive sparse sensing algorithm; Step 2: Construct the GraphSAGE network model, including the GraphSAGE network and a fully connected classifier. The GraphSAGE network is divided into two parts: neighbor node feature aggregation and target node feature update. Step 3: Model training phase; Train the network model using labeled graph structured data, and introduce the cross-entropy function as the loss function to evaluate model performance; Step 4: Classification stage.

[0005] Preferably, step 1 specifically comprises: Step 1-1: Processing multi-function radar PDW data X , Normalization is performed to obtain the normalized PDW dataset. The normalization method used is maximum-minimum normalization; the specific formula is:

[0006] in Indicates the first pulse sequence of the radiation source The first PDW data in Characteristic parameters, The dimension of the feature parameters, The length of the radiation source pulse sequence, Represents the normalized result ; Steps 1-2: Treat each normalized PDW data point as a graph node, and use Gaussian kernel density estimation to calculate the... i The spatial density of a node is given by the following formula:

[0007] in Indicates the first i Spatial density of nodes Indicates kernel bandwidth. Representing nodes respectively i , j eigenvectors; Steps 1-3: For each node i The number of neighborhood connections is adaptively determined based on its normalized spatial density. k The specific calculation formula is as follows:

[0008] in Indicates the first i The number of neighborhood connections of each node. Indicates the preset neighborhood connection cardinality. This represents the neighborhood connectivity adjustment coefficient; Steps 1-4: Based on the number of neighborhood connections of each node , find Find the nearest neighboring nodes, define the relationship between them, and calculate the weight of the edge between the nodes. The specific calculation formula is as follows:

[0009] in, Represents a node i , j The weight of the edges connecting them; Let the elements at the corresponding positions in the adjacency matrix A This led to the construction of a weighted undirected graph structure data. ,in V This represents the set of nodes in the graph. E This represents the set of edges in the graph.

[0010] Preferably, step 2 specifically comprises: Step 2-1: Neighbor node feature aggregation; For each node, a fixed number of nodes are sampled from its neighbors to form the sampling neighborhood. The features of these neighboring nodes are weighted and aggregated, and the specific formula is as follows:

[0011] in Indicates the first l Layer nodes i The weighted aggregation result of neighbor node features For nodes i The set of sampled neighbor nodes, For neighbor node j in the th... Features of the layer; Step 2-2: Target node feature update; The GraphSAGE network combines the original features of the target node with the aggregated features of its neighbors and updates the node features using a non-linear activation function. The specific formula is as follows:

[0012] in Indicates the first l Layer nodes i Features Indicates the first l The trainable weight matrix of the layer, This represents the concatenation of two vectors. This represents the activation function. Indicates the first l-1 Layer nodes i Features; Steps 2-3: For a L In a multi-layered GraphSAGE network, the final feature representation of each node is as follows: This feature includes the node's own features and the features of its local neighboring nodes; a fully connected layer is used to perform a non-linear mapping on the node features, and the probability distribution of the node's category is output through the Softmax activation function, specifically expressed as follows:

[0013] in Represents a node i The probability of belonging to each category. This represents the weight matrix of the fully connected layer.

[0014] Preferably, step 3 specifically comprises: The cross-entropy function is specifically expressed as:

[0015] in, Represents a set of labeled nodes. C For the number of categories, One-hot encoding of the real label. For nodes i Category c The probability of; Backpropagation is used to optimize the parameters of each layer of the GraphSAGE network and the parameters of the fully connected layers until the loss function converges, thus completing the model training.

[0016] Preferably, step 4 specifically comprises: After model training is complete, for each test node, the final feature representation of the node is first generated using the trained GraphSAGE network. The data is then fed into a fully connected layer to obtain the predicted probability for each category. The category label corresponding to the highest probability is used as the prediction result; thus, the classification of graph nodes is completed, and radar signal sorting is realized.

[0017] An electronic device includes: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the electronic device to perform the radar signal sorting method described above.

[0018] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the radar signal sorting method described above.

[0019] A chip includes a processor for calling and running a computer program from a memory, causing a device on which the chip is installed to perform the radar signal sorting method described above.

[0020] A computer program product includes a computer storage medium storing a computer program, the computer program including instructions executable by at least one processor, which, when executed by the at least one processor, implement the radar signal sorting method described above.

[0021] The beneficial effects of this invention are as follows: 1. This invention proposes an adaptive sparse sensing algorithm to construct radar PDW data into graph structure data, which can effectively reduce redundant connections between nodes in high-density areas, while enhancing the topological association between nodes in low-density areas, thereby improving the expressive power of the graph structure. 2. This invention proposes a weighted undirected graph construction method based on Gaussian kernel function, which introduces edge weight information to reflect the similarity relationship between nodes, enhances the graph structure data information, and helps to improve the accuracy of radar signal sorting; 3. This invention employs a GraphSAGE network model based on a weighted aggregation mechanism, which can effectively integrate edge weight information to aggregate neighbor features, thereby classifying nodes in complex graph structure data and improving the accuracy of radar signal sorting. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a sample distribution diagram of radar simulation parameters in the feature space in an embodiment of the present invention; Figure 3 shows the sorting results of four different methods in the embodiments of the present invention. (a) is the sorting result of the k-means algorithm, (b) is the sorting result of the DBSCAN algorithm, (c) is the sorting result of the GCN-ResNet algorithm, and (d) is the sorting result of the method of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] This invention provides a radar signal sorting technology based on adaptive sparse sensing and GraphSAGE, which is used to solve the problem of insufficient sorting performance of reconnaissance receivers in complex electromagnetic environments when facing high-density and unevenly distributed radiation source pulses.

[0025] To achieve the above objectives, the technical solution adopted by the present invention mainly includes the following steps: Step 1: Graph Structure Data Construction Stage. The graph structure data is constructed using an adaptive sparse sensing algorithm. The specific steps are as follows: Step 1-1: Processing multi-function radar PDW data X ( The PDW dataset is then normalized to obtain the normalized dataset. The normalization method used is maximum-minimum normalization. The specific formula is:

[0026] in Indicates the first pulse sequence of the radiation source The first PDW data in Characteristic parameters, The dimension of the feature parameters, The length of the radiation source pulse sequence, Represents the normalized result .

[0027] Steps 1-2: Treat each normalized PDW data point as a graph node, and use Gaussian kernel density estimation to calculate the... i The spatial density of a node is given by the following formula:

[0028] in Indicates the first i Spatial density of nodes Indicates kernel bandwidth. Represents a node i , j eigenvectors.

[0029] Steps 1-3: For each node i The number of neighborhood connections is adaptively determined based on its normalized spatial density. k The specific calculation formula is as follows:

[0030] in Indicates the first i The number of neighborhood connections of each node. Indicates the preset neighborhood connection cardinality. This represents the neighborhood connectivity adjustment coefficient.

[0031] Steps 1-4: Based on the number of neighborhood connections of each node , find Find the nearest neighboring nodes, define the relationship between them, and calculate the weight of the edge between the nodes. The specific calculation formula is as follows:

[0032] in, Represents a node i , j The weight of the edges connecting them.

[0033] Let the elements at the corresponding positions in the adjacency matrix A This leads to the construction of a weighted undirected graph structure data. ,in V This represents the set of nodes in the graph. E This represents the set of edges in the graph.

[0034] Step 2: Construct the GraphSAGE network model, which mainly includes the GraphSAGE network and a fully connected classifier. The GraphSAGE network is mainly divided into two parts: neighbor node feature aggregation and target node feature update, specifically: Step 2-1: Neighbor Node Feature Aggregation. For each node, a fixed number of nodes are sampled from its neighbors to form a sampling neighborhood. The features of these neighboring nodes are weighted and aggregated, and the specific formula is as follows:

[0035] in Indicates the first l Layer nodes i The weighted aggregation result of neighbor node features For nodes i The set of sampled neighbor nodes, For neighbor node j in the th... Characteristics of the layer.

[0036] Step 2-2: Target Node Feature Update. The GraphSAGE network combines the original features of the target node with the aggregated neighbor features and updates the node features using a non-linear activation function. The specific formula is:

[0037] in Indicates the first l Layer nodes i Features Indicates the first l The trainable weight matrix of the layer, This represents the concatenation of two vectors. This represents the activation function.

[0038] Steps 2-3: For a L In a multi-layered GraphSAGE network, the final feature representation of each node is as follows: This feature includes the node's own features as well as the features of its local neighboring nodes. To complete the classification task, a fully connected layer is used to perform a non-linear mapping on the node features, and the probability distribution of the node's category is output through the Softmax activation function, specifically represented as follows:

[0039] in Represents a node i The probability of belonging to each category. This represents the weight matrix of the fully connected layer.

[0040] Step 3: Model Training Phase. The network model is trained using labeled graph-structured data. The cross-entropy function is introduced as the loss function to evaluate model performance, specifically expressed as:

[0041] in, Represents a set of labeled nodes. C For the number of categories, One-hot encoding of the real label. For nodes i Category c The probability of.

[0042] Backpropagation is used to optimize the parameters of each layer of the GraphSAGE network and the parameters of the fully connected layers until the loss function converges, thus completing the model training.

[0043] Step 4: Classification Stage. After model training is complete, for each test node, the final feature representation of the node is first generated using the trained GraphSAGE network. The data is then fed into a fully connected layer to obtain the predicted probability for each category. The category label corresponding to the highest probability is used as the prediction result. This completes the classification of graph nodes and enables radar signal sorting.

[0044] Example: To verify the effectiveness of the proposed algorithm, this example designed a PDW parameter simulation experiment involving four radars, with parameters shown in Table 1. Specifically, two conventional radars and two multi-functional radars were simulated, incorporating four parameters: PW, RF, DOA, and PRI, and featuring multiple modulation schemes. To better reflect the real environment, the parameter settings of the four radars overlapped to some extent, and the number of pulses for each radar was randomized. It should be noted that PRI is only used to count the number of pulses within the simulation time and is difficult to accurately estimate in actual measurements. Therefore, PW, RF, and DOA were selected as the input pulse signal. Figure 2The distribution of simulation parameters of four radars in the three-dimensional feature space is shown.

[0045] Table 1 Radar signal simulation parameters

[0046] The multifunctional radar signal sorting method proposed in this invention consists of four stages: graph structure data construction, GraphSAGE network model construction, model training, and graph node classification. The specific implementation steps are as follows: Step 1: Graph Structure Data Construction Stage. Graph structure data is constructed based on the received multi-function radar PDW data X, where the PDW data is represented as... ,in , They represent the first i The three characteristic parameters of each PDW data are pulse width, frequency, and angle of arrival; N The length of the radiation source pulse sequence, i.e., the amount of PDW data, is determined by the following steps: Step 1-1: Processing multi-function radar PDW data X Normalization is performed to obtain the normalized PDW dataset. The normalization method used is maximum-minimum normalization. The specific formula is:

[0047] in Indicates the first pulse sequence of the radiation source The first PDW data in Characteristic parameters, The dimension of the feature parameter is 3 in this embodiment. The length of the radiation source pulse sequence is 10000 in this embodiment. Represents the normalized result .

[0048] Steps 1-2: Treat each normalized PDW data point as a graph node, and use Gaussian kernel density estimation to calculate the... i The spatial density of a node is given by the following formula:

[0049] in Indicates the first i Spatial density of nodes This represents the kernel bandwidth, which is 0.1 in this embodiment. Represents a node i , j eigenvectors.

[0050] Steps 1-3: For each node i The number of neighborhood connections is adaptively determined based on its spatial density. k The specific calculation formula is as follows:

[0051] in Indicates the first i The number of neighborhood connections of each node. This indicates the preset neighborhood connection cardinality, which is 10 in this embodiment. This represents the neighborhood connection number adjustment coefficient, which is 10 in this embodiment.

[0052] Steps 1-4: Based on the number of neighborhood connections of each node , find Find the nearest neighboring nodes, define the relationship between them, and calculate the weight of the edge between the nodes. The specific calculation formula is as follows:

[0053] in, Represents a node i , j The weight of the edges connecting them.

[0054] Let the elements at the corresponding positions in the adjacency matrix A This leads to the construction of a weighted undirected graph structure data. ,in V This represents the set of nodes in the graph. E This represents the set of edges in the graph.

[0055] Step 2: GraphSAGE Network Model Construction Stage. This mainly includes the GraphSAGE network and a fully connected classifier. The GraphSAGE network is primarily divided into two parts: neighbor node feature aggregation and target node feature update. Specifically: Step 2-1: Neighbor Node Feature Aggregation. For each node, a fixed number of nodes are sampled from its neighbors to form a sampling neighborhood. The features of these neighboring nodes are weighted and aggregated, and the specific formula is as follows:

[0056] in Indicates the first l Layer nodes i The weighted aggregation result of neighbor node features For nodes i The set of sampled neighbor nodes, For neighbor node j in the th... Characteristics of the layer.

[0057] Step 2-2: Target Node Feature Update. The GraphSAGE network combines the original features of the target node with the aggregated neighbor features and updates the node features using a non-linear activation function. The specific formula is as follows:

[0058] in Indicates the first l Layer nodes i Features Indicates the first l The trainable weight matrix of the layer, This represents the concatenation of two vectors. This represents the activation function.

[0059] Steps 2-3: For a L In a multi-layered GraphSAGE network, the final feature representation of each node is as follows: This feature includes the node's own features as well as the features of its local neighboring nodes. To complete the classification task, a fully connected layer is used to perform a non-linear mapping on the node features, and the probability distribution of the node's category is output through the Softmax activation function, specifically represented as follows:

[0060] in Represents a node i The probability of belonging to each category. This represents the weight matrix of the fully connected layer.

[0061] Step 3: Model Training Phase. The network model is trained using labeled graph-structured data. The cross-entropy function is introduced as the loss function to evaluate model performance, specifically expressed as:

[0062] in, Represents a set of labeled nodes. C For the number of categories, One-hot encoding of the real label. For nodes i Category c The probability of.

[0063] The backpropagation algorithm is used to optimize the parameters of each layer of the GraphSAGE network and the parameters of the fully connected layer until the loss function converges, thus completing the model training.

[0064] Step 4: Graph Node Classification Stage. After model training is complete, for each test node, the final feature representation of the node is first generated using the trained GraphSAGE network. The data is then fed into a fully connected layer to obtain the predicted probability for each category. The category label corresponding to the highest probability is used as the prediction result. This completes the classification of graph nodes and enables radar signal sorting.

[0065] Based on the comparison between this method and the k-means clustering algorithm, the DBSACN algorithm, and radar signal sorting algorithms based on GCN and ResNet, the parameters set for the four methods are as follows: 1. k-means algorithm: The initial number of clusters is set to 4, and the maximum number of iterations is set to 100; 2. DBSCAN algorithm: The neighborhood radius is set to 0.08, and the minimum number of sample points in the neighborhood is set to 100; 3. GCN-ResNet algorithm: initial learning rate 0.01, weight decay value 0.0005, training epochs 500, three graph convolutional layers, output dimensions 16-16-4, test set to training set ratio 9:1; 4. The method of this invention: initial learning rate 0.02, weight decay value 0.0005, training epochs 500, three graphsage layers, output dimensions 16-32-16 respectively, and the ratio of test set to training set is 9:1; The sorting results of the method of this invention and the other three algorithms on the radar PDW dataset are shown in Figure 3. Among them, Figure 3(a) shows the sorting result of the k-means algorithm, Figure 3(b) shows the sorting result of the DBSCAN algorithm, Figure 3(c) shows the sorting result of the C-Net algorithm, and Figure 3(d) shows the sorting result of the method of this invention.

[0066] Analysis of Figure 3 reveals two shortcomings of the k-means algorithm: firstly, it requires specifying the initial number of clusters and lacks adaptability; secondly, it cannot effectively handle clusters of arbitrary shapes, resulting in poor sorting performance when there is overlap in the parameters of multi-functional radars. The DBSCAN algorithm is significantly affected by the distribution of the sample space. In this experiment, due to severe overlap in the feature parameter space of the four radars, the algorithm mistakenly divided the four radars into two: radar 1 was sorted as radar 4, some pulses from radar 2 were sorted as radar 3, and the remaining pulses were treated as noise and discarded. The GCN-ResNet algorithm failed to fully learn all the data information, resulting in poor sorting performance for radars 2 and 3, and misclassifying some pulses as radar 4. In contrast, the method proposed in this invention can correctly sort all four radars, with poor sorting performance only at the overlap between radar 2 and radars 3 and 4, leading to a lower sorting accuracy for radar 2 compared to the other radars. Overall, this method outperforms the other three methods in terms of sorting accuracy.

[0067] This experiment uses sorting accuracy as the evaluation metric to measure the sorting performance of four sorting algorithms on the radar PDW simulation dataset. The results are shown in Table 2. The sorting accuracy rates of the four algorithms are 70.91%, 64.77%, 83.57%, and 93.57%, respectively. It is evident that the sorting accuracy of the method described in this invention is significantly better than the other algorithms. With a labeling rate of 0.1, the accuracy is improved by 9.74% compared to the GCN-ResNet algorithm. This is because the method in this invention, when constructing graph-structured data, can adaptively connect neighboring nodes according to different densities and introduce weighted coefficients. Combined with a graphsage network based on a weighted aggregation mechanism, this effectively enhances the feature representation capability, enabling each node to more accurately fuse information from its neighboring nodes, thereby improving the radar signal sorting effect.

[0068] Table 2. Comparison of sorting performance of the four algorithms

Claims

1. A radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation, characterized in that, Includes the following steps: Step 1: Construct graph-structured data using an adaptive sparse sensing algorithm; Step 2: Construct the GraphSAGE network model, including the GraphSAGE network and a fully connected classifier. The GraphSAGE network is divided into two parts: neighbor node feature aggregation and target node feature update. Step 3: Model training phase; Train the network model using labeled graph structured data, and introduce the cross-entropy function as the loss function to evaluate model performance; Step 4: Classification stage.

2. The radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation according to claim 1, characterized in that, Step 1 specifically involves: Step 1-1: Processing multi-function radar PDW data X , Normalization is performed to obtain the normalized PDW dataset. The normalization method used is maximum-minimum normalization; the specific formula is: in Indicates the first pulse sequence of the radiation source The first PDW data in Characteristic parameters, The dimension of the feature parameters, The length of the radiation source pulse sequence, Represents the normalized result ; Steps 1-2: Treat each normalized PDW data point as a graph node, and use Gaussian kernel density estimation to calculate the... i The spatial density of a node is given by the following formula: in Indicates the first i Spatial density of nodes Indicates kernel bandwidth. Representing nodes respectively i , j eigenvectors; Steps 1-3: For each node i The number of neighborhood connections is adaptively determined based on its normalized spatial density. k The specific calculation formula is as follows: in Indicates the first i The number of neighborhood connections of each node. Indicates the preset neighborhood connection cardinality. This represents the neighborhood connectivity adjustment coefficient; Steps 1-4: Based on the number of neighborhood connections of each node , find Find the nearest neighboring nodes, define the relationship between them, and calculate the weight of the edge between the nodes. The specific calculation formula is as follows: in, Represents a node i , j The weight of the edges connecting them; Let the elements at the corresponding positions in the adjacency matrix A This led to the construction of a weighted undirected graph structure data. ,in V This represents the set of nodes in the graph. E This represents the set of edges in the graph.

3. The radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation according to claim 2, characterized in that, Step 2 specifically involves: Step 2-1: Neighbor node feature aggregation; For each node, a fixed number of nodes are sampled from its neighbors to form the sampling neighborhood. The features of these neighboring nodes are weighted and aggregated, and the specific formula is as follows: in Indicates the first l Layer nodes i The weighted aggregation result of neighbor node features For nodes i The set of sampled neighbor nodes, For neighbor node j in the th... Characteristics of the layer; Step 2-2: Target node feature update; The GraphSAGE network combines the original features of the target node with the aggregated features of its neighbors and updates the node features using a non-linear activation function. The specific formula is as follows: in Indicates the first l Layer nodes i Features Indicates the first l The trainable weight matrix of the layer, This represents the concatenation of two vectors. This represents the activation function. Indicates the first l-1 Layer nodes i Features; Steps 2-3: For a L In a multi-layered GraphSAGE network, the final feature representation of each node is as follows: This feature includes the node's own features and the features of its local neighboring nodes; a fully connected layer is used to perform a non-linear mapping on the node features, and the probability distribution of the node's category is output through the Softmax activation function, specifically expressed as follows: in Represents a node i The probability of belonging to each category. This represents the weight matrix of the fully connected layer.

4. The radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation according to claim 3, characterized in that, Step 3 specifically involves: The cross-entropy function is specifically expressed as: in, Represents a set of labeled nodes. C For the number of categories, One-hot encoding of the real label. For nodes i Category c The probability of; Backpropagation is used to optimize the parameters of each layer of the GraphSAGE network and the parameters of the fully connected layers until the loss function converges, thus completing the model training.

5. The radar signal sorting method based on adaptive sparse sensing and graph sampling aggregation according to claim 4, characterized in that, Step 4 specifically involves: After model training is complete, for each test node, the final feature representation of the node is first generated using the trained GraphSAGE network. The data is then fed into a fully connected layer to obtain the predicted probability for each category. The category label corresponding to the highest probability is used as the prediction result; thus, the classification of graph nodes is completed, and radar signal sorting is realized.

6. An electronic device, characterized in that, include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 5.

8. A chip, characterized in that, include: A processor for retrieving and running a computer program from memory, causing a device on which the chip is mounted to perform the method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product includes a computer storage medium storing a computer program, the computer program including instructions executable by at least one processor, which, when executed by the at least one processor, implement the method as described in any one of claims 1 to 5.

Images