Multi-dimensional information event identification method and system based on multi-core support vector machine

Abstract

The invention discloses a multi-dimensional information event identification method and system based on a multi-kernel support vector machine, and the method comprises the steps: extracting multi-dimensional feature vectors from a power grid, the multi-dimensional feature vectors comprise a part or all of time domain partial feature values and frequency domain partial feature values, the time domain partial feature values comprise effective values, skewness and kurtosis of voltage and current, and the frequency domain partial feature values comprise effective values, skewness and kurtosis of frequency domain; the characteristic values of the frequency domain part comprise amplitude spectrums of voltage and current, phase spectrums and specified hth harmonic amplitudes; the multi-dimensional feature vectors are input into a multi-kernel support vector machine to achieve multi-dimensional information event recognition, and a kernel function of the multi-kernel support vector machine is obtained through weighted fusion of kernel functions of a Gaussian kernel and a quantum kernel; and the quantum kernel is a square module of an inner product between corresponding quantum states of different feature vectors in the multi-dimensional feature vectors and is used for quantifying the global structure similarity of the feature vectors. The invention aims to overcome the defect of insufficient event recognition performance of a power system under the condition of small samples in a traditional method, and improves the robustness and accuracy of event recognition.

Description

Technical Field

[0001] This invention belongs to the field of power grid disturbance identification, specifically relating to a multidimensional information event identification method and system based on multi-kernel support vector machine. Background Technology

[0002] With the increasing integration of new energy power generation systems into the power grid, inverter resources are playing an increasingly important role in the power system. Against this backdrop, disturbances caused by inverters, such as voltage fluctuations, frequency drops, and waveform distortions, pose significant challenges to event detection and classification. Notably, these inverter-generated disturbances can easily trigger chain reactions, further threatening the stability of power system voltage and frequency. Therefore, identifying event types and eliminating them promptly is crucial. Neural network-based event recognition methods rely on large amounts of data for event identification. Examples include multi-task time-frequency transform networks (MTNs) for identifying abnormal events such as power generation output, phase shifts, and load shedding; and GNNs that integrate physical information and measurement data, utilizing system impedance and network topology to identify anomalies. However, these methods cannot meet the requirements for event recognition under small sample conditions. While traditional support vector machines (SVMs) perform well in small sample event recognition, their use of a single kernel function, such as a Gaussian kernel, has significant limitations, including insufficient generalization ability for small samples, inadequate utilization of multi-dimensional features, and low sensitivity to global structure. Therefore, methods that meet the requirements of accuracy and robustness in event recognition under small sample conditions are essential. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a multidimensional information event recognition method and system based on multi-kernel support vector machines, which addresses the above-mentioned problems of the prior art. This invention aims to overcome the insufficient performance of traditional methods in event recognition in power systems with small sample sizes, and improve the robustness and accuracy of event recognition.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multidimensional information event recognition method based on multi-kernel support vector machines includes the following steps: S101, extract multi-dimensional feature vectors from the power grid, including some or all of the time-domain feature values ​​and frequency-domain feature values. The time-domain feature values ​​include some or all of the effective values, skewness, and kurtosis of voltage and current. The frequency-domain feature values ​​include some or all of the amplitude spectrum, phase spectrum, and specified h-th harmonic amplitude of voltage and current. S102, input the multidimensional feature vector into the multi-kernel support vector machine to realize multidimensional information event recognition. The kernel function of the multi-kernel support vector machine is obtained by weighted fusion of the kernel functions of Gaussian kernel and quantum kernel. The quantum kernel is the square modulus of the inner product between the corresponding quantum states of different feature vectors in the multidimensional feature vector to quantize the global structural similarity of the feature vector.

[0005] Optionally, the functional expression of the kernel function of the quantum nucleus is: , in, For the kernel function of the quantum nucleus, and These are multidimensional feature vectors. The i-th eigenvector and the j-th eigenvector in the image. Representing the eigenvector quantum state and quantum state The inner product between For all qubits, the measurement is as follows The probability, For feature vectors The conjugate of quantum feature maps, For feature vectors Quantum feature map, for The initial state of each qubit.

[0006] Optionally, the kernel function of the multi-kernel support vector machine is expressed as follows: , , in, For the kernel function of a multi-kernel support vector machine, and These are the kernel functions for Gaussian kernels used to process time-domain features and frequency-domain features, respectively. and These are the kernel functions for quantum kernels that process time-domain and frequency-domain features, respectively. , , , , , It is the weighting coefficient.

[0007] Optionally, the function expression for inputting the multidimensional feature vector into a multi-kernel support vector machine to achieve multidimensional information event recognition is: , in, This represents the multidimensional information event recognition results from a multi-kernel support vector machine. For symbolic functions, For the weight vector, the superscript... For transpose operation, For the kernel function of a multi-kernel support vector machine, This is a bias term.

[0008] Optionally, the training of the multi-kernel support vector machine includes: S201, extract multidimensional feature vectors and their labels from multidimensional information event samples from the power grid to construct a training dataset; S202, Adjust weighting coefficients , , , , , And construct the kernel function for the multi-kernel support vector machine; S203, train a multi-kernel support vector machine using the training dataset and optimize the model parameters according to the following formula: , , , in, For the weight vector, For bias terms, is the penalty coefficient in the kernel function of the Gaussian kernel. The number of samples in the training dataset, For the first Slack variables for each sample, For the first The label of each sample The kernel function is the kernel function of the multi-kernel support vector machine; the model parameters include the weight coefficients in the kernel function of the multi-kernel support vector machine. , , , , , And the bias term in the kernel function of the Gaussian kernel. and penalty coefficient ; S204, determine whether the preset termination condition has been met. If the preset termination condition has not been met, jump to step S202 to continue training; otherwise, end and exit.

[0009] Optionally, the function expression for the preset termination condition in step S204 is: , in, For prediction error, , For the new model parameters, For the current model parameters, This represents a multi-kernel support vector machine.

[0010] Optionally, the kernel function of the Gaussian kernel can be expressed as follows: , in, Let be the kernel function of the Gaussian kernel. and These are multidimensional feature vectors. The first in 3D eigenvectors and the 1st eigenvector 3D feature vectors is the width of the Gaussian kernel.

[0011] Furthermore, this embodiment also provides a multidimensional information event recognition system based on a multi-core support vector machine, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the multidimensional information event recognition method based on the multi-core support vector machine.

[0012] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program or instructions that are programmed or configured to execute the multidimensional information event recognition method based on a multi-core support vector machine via a processor.

[0013] Furthermore, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the multidimensional information event recognition method based on a multi-core support vector machine via a processor.

[0014] Compared with existing technologies, the present invention mainly achieves the following beneficial effects: The method of the present invention includes extracting multidimensional feature vectors from the power grid, including some or all of the time-domain and frequency-domain feature values. The time-domain feature values ​​include the effective values, skewness, and kurtosis of voltage and current, while the frequency-domain feature values ​​include the amplitude spectrum, phase spectrum, and specified h-th harmonic amplitude of voltage and current. The multidimensional feature vectors are input into a multi-kernel support vector machine to achieve multidimensional information event recognition. The kernel function of the multi-kernel support vector machine is obtained by weighted fusion of the kernel functions of Gaussian kernels and quantum kernels. The quantum kernel is the square modulus of the inner product between the corresponding quantum states of different feature vectors within the multidimensional feature vector, used to quantify the global structural similarity of the feature vectors. Compared with traditional classification methods, the advantage of the method of the present invention is that it can use multiple kernel functions to map each input feature, thereby improving classification accuracy. It also has high accuracy under small sample conditions, effectively overcoming the insufficient performance of traditional methods in event recognition of power systems under small sample conditions, and improving the robustness and accuracy of event recognition. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the training process of the multi-kernel support vector machine in an embodiment of the present invention.

[0016] Figure 2 This is a performance diagram illustrating different training data ratios in an embodiment of the present invention.

[0017] Figure 3 This is a schematic diagram comparing the performance of different event recognition methods in embodiments of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] like Figure 1 As shown, the multidimensional information event recognition method based on multi-kernel support vector machines in this embodiment includes the following steps: S101, extract multi-dimensional feature vectors from the power grid, including some or all of the time-domain feature values ​​and frequency-domain feature values. The time-domain feature values ​​include some or all of the effective values, skewness, and kurtosis of voltage and current. The frequency-domain feature values ​​include some or all of the amplitude spectrum, phase spectrum, and specified h-th harmonic amplitude of voltage and current. S102, input the multidimensional feature vector into the multi-kernel support vector machine (QMSVM) to realize multidimensional information event recognition. The kernel function of the multi-kernel support vector machine is obtained by weighted fusion of the kernel functions of Gaussian kernel and quantum kernel. The quantum kernel is the square modulus of the inner product between the corresponding quantum states of different feature vectors in the multidimensional feature vector to quantize the global structural similarity of the feature vector.

[0020] In this embodiment, the kernel function of the Gaussian kernel is expressed as follows: , in, Let be the kernel function of the Gaussian kernel. and These are multidimensional feature vectors. The first in 3D eigenvectors and the 1st eigenvector 3D feature vectors The width of the Gaussian kernel. , used to control the mapping results.

[0021] The intrinsic properties of kernel functions significantly impact recognition performance. Gaussian kernels primarily rely on the Euclidean distance between samples, exhibiting strong local learning capabilities. However, their sensitivity to the global structure of the data is limited. Therefore, using only Gaussian kernels for event recognition based on multidimensional information of event samples can result in relatively large errors. Introducing quantum kernels to increase sensitivity to the global data can improve the accuracy of event recognition. This involves transforming the dimensional feature vector... The method for converting to quantum states for computation is as follows: convert the multidimensional eigenvectors Mapping to quantum state ,in The quantum superposition representation of the features is achieved by performing quantum gate operations on the initial all-zero state through a quantum feature map (quantum circuit): , in, Indicates an initial all-zero state. For feature vectors The quantum feature map. With the input being... and In this case, the quantum kernel is defined as the square modulus of the inner product between its corresponding quantum states, quantizing the global structural similarity of the eigenvectors. In this embodiment, the kernel function of the quantum kernel is expressed as: , in, For the kernel function of the quantum nucleus, and These are multidimensional feature vectors. The i-th eigenvector and the j-th eigenvector in the image. Representing the eigenvector quantum state and quantum state The inner product between For all qubits, the measurement is as follows The probability, For feature vectors The conjugate of quantum feature maps, For feature vectors Quantum feature map, for The initial state of 1 qubit; It is the symbol for tensor product; Indicates to Perform n tensor product operations, that is, n tensors Combination of states. Quantum kernels utilize the properties of quantum entanglement to capture higher-order correlations between features, compensating for the Gaussian kernel's deficiency in perceiving the global structure. Based on the above, Gaussian kernels and quantum kernels are combined into multiple kernels. It should be noted that quantum feature maps (quantum circuits) are existing known methods, and for details, please refer to: Gong Jing, He Min, Yao Zeqing. Realization of quantum circuits by entanglement swapping [J]. Communication Technology, 2008, 41(6):3.DOI:10.3969 / j.issn.1002-0802.2008.06.027.

[0022] The weighted fusion of Gaussian and quantum kernel kernels into a multi-kernel system can be achieved using any desired fusion method. For example, as an optional implementation, the kernel function of the multi-kernel support vector machine in this embodiment is expressed as follows: , , in, For the kernel function of a multi-kernel support vector machine, and These are the kernel functions for Gaussian kernels used to process time-domain features and frequency-domain features, respectively. and These are the kernel functions for quantum kernels that process time-domain and frequency-domain features, respectively. , , , , , These are weighting coefficients. The weights for time-domain features, The weights are the frequency domain features. and These represent the weights of the Gaussian kernel in the time domain and frequency domain features, respectively. and These represent the weights of the quantum kernel component in the time and frequency domains, respectively. Novel events exhibit highly nonlinear relationships due to environmental influences and change rapidly over time. At the microscopic level, quantum effects in power electronic devices can affect the dynamic behavior of the system. These features exhibit complex quantum correlations in high-dimensional space, and traditional kernel functions lack sufficient feature representation capabilities. In this embodiment, the kernel function of the multi-kernel support vector machine adaptively adjusts its weights, enabling differentiated modeling of features across different dimensions and improving the expressive power of mixed features.

[0023] In this embodiment, the function expression for inputting multidimensional feature vectors into a multi-kernel support vector machine to achieve multidimensional information event recognition is as follows: , in, This represents the multidimensional information event recognition results from a multi-kernel support vector machine. For symbolic functions, For the weight vector, the superscript... For transpose operation, For the kernel function of a multi-kernel support vector machine, This is a bias term.

[0024] like Figure 1 As shown, the training of the multi-kernel support vector machine in this embodiment includes: S201, Extract multidimensional feature vectors and their labels from multidimensional information event samples from the power grid to construct a training dataset, multidimensional information event samples. It can be represented as: , in, These are multidimensional feature vectors. The first in 3D feature vectors and their labels The number of dimensions; S202, Adjust weighting coefficients , , , , , And construct the kernel function for the multi-kernel support vector machine (QMSVM); S203, In order to learn the decision plane for classification, a multi-kernel support vector machine (QMSVM) is trained using the training dataset, and the model parameters are optimized according to the following formula: , , , in, For the weight vector, For bias terms, is the penalty coefficient in the kernel function of the Gaussian kernel (used to balance marginal maximization and classification error). The number of samples in the training dataset, For the first Slack variables for each sample, For the first The label of each sample The kernel function of the multi-kernel support vector machine is given above. By introducing the dual Lagrangian function, the model parameters of the proposed QMSVM can be obtained through partial derivatives. These model parameters include the weight coefficients in the kernel function of the multi-kernel support vector machine. , , , , , And the bias term in the kernel function of the Gaussian kernel. and penalty coefficient ; S204, determine whether the preset termination condition has been met. If the preset termination condition has not been met, jump to step S202 to continue training; otherwise, end and exit.

[0025] like Figure 1 As shown, the function expression for the preset termination condition in step S204 is: , in, For prediction error, , For the new model parameters, For the current model parameters, This represents a multi-kernel support vector machine. The training of a multi-kernel support vector machine can use simulated annealing to optimize parameters and improve recognition performance. By using the aforementioned preset termination condition, the "temperature" parameter can be gradually reduced to narrow the search range until it converges to the globally optimal parameters.

[0026] To verify the multi-core support vector machine (QMSVM) of this embodiment, a distributed simulation model containing 14 photovoltaic (PV) nodes was established in OpenDSS and ATP-EMTP based on the IEEE 123 bus test feeder. Load conditions, PV penetration levels, and node locations were specifically configured to simulate common measurement uncertainties in real power systems. Furthermore, to test the robustness of the QMSVM under various operating conditions, the simulation process diversified load levels, fault locations, and switching phases, simulating typical disturbances such as three-phase, single-phase, and line-to-line faults. To verify the robustness of the proposed QMSVM, the performance of three existing support vector machines—Sigmoid kernel SVM (SSVM), Gaussian kernel SVM (GSVM), and Quantum kernel SVM (QSVM)—was compared with that of the proposed QMSVM under different training data ratios. This means that different numbers of training samples were used during training, and the final detection results are as follows: Figure 2 As shown, n% of the training data represents n% of the data randomly selected from 3767 samples. Figure 2 As can be seen, across all training data proportions, the classification accuracy of the proposed multi-kernel support vector machine consistently outperforms other support vector machine methods. Notably, the accuracy of Gaussian kernel SVM and Quantum kernel SVM is slightly lower than that of the proposed multi-kernel support vector machine. This indicates that both global and local information contribute to anomaly detection. The key difference lies in the careful adjustment of the weight coefficients to balance the impact of the two tasks. Furthermore, as the training sample proportion increases, the accuracy of all methods improves accordingly, with the proposed multi-kernel support vector machine consistently achieving the highest accuracy. Specifically, at training ratios of 40% and 70%, the accuracy of the proposed multi-kernel support vector machine reaches 92.44% and 93.97%, respectively. This demonstrates that the proposed multi-kernel support vector machine exhibits good stability and robustness even with small sample sizes. To further verify the performance of the multi-kernel support vector machine proposed in this embodiment, five mainstream machine learning methods were selected for comparison: decision tree (DT), Naive Bayes (NB), K-nearest neighbor (KNN), gradient boosting decision tree (GBDT), and support vector machine (SVM). These methods were used to identify multi-dimensional event types, and the performance of each method was comprehensively evaluated using four metrics: precision, recall, and F1 score. The comparison results are as follows: Figure 3 As shown. From Figure 3As can be seen, the multi-kernel support vector machine proposed in this embodiment outperforms other machine learning models on all evaluation metrics. Specifically, the multi-kernel support vector machine proposed in this embodiment achieves over 94% in precision, recall, and F1 score, significantly surpassing gradient boosting decision trees, Naive Bayes, decision trees, and K-nearest neighbors. Furthermore, compared to traditional support vector machines, the multi-kernel support vector machine proposed in this embodiment exhibits higher precision and F1 score. These results demonstrate the superior performance of the multi-kernel support vector machine proposed in this embodiment in event type detection.

[0027] In summary, the method of this embodiment includes extracting multidimensional feature vectors from the power grid, including some or all of the time-domain and frequency-domain feature values. The time-domain feature values ​​include the effective values, skewness, and kurtosis of voltage and current, while the frequency-domain feature values ​​include the amplitude spectrum, phase spectrum, and specified h-th harmonic amplitude of voltage and current. The multidimensional feature vectors are then input into a multi-kernel support vector machine (MSVM) to achieve multidimensional information event recognition. The kernel function of the MSVM is obtained by weighted fusion of Gaussian and quantum kernels. The quantum kernel is the square modulus of the inner product between corresponding quantum states of different feature vectors within the multidimensional feature vector, used to quantify the global structural similarity of the feature vectors. Compared to traditional classification methods, the advantage of this embodiment is that it can use multiple kernel functions to map each input feature, thereby improving classification accuracy. It also has high accuracy under small sample conditions, eliminating the time-consuming training with a large number of samples and the difficulty of sample collection. It effectively overcomes the insufficient performance of traditional methods in power system event recognition under small sample conditions, improving the robustness and accuracy of event recognition.

[0028] Furthermore, this embodiment also provides a multidimensional information event recognition system based on a multi-core support vector machine, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the multidimensional information event recognition method based on the multi-core support vector machine.

[0029] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program or instructions that are programmed or configured to execute the multidimensional information event recognition method based on a multi-core support vector machine via a processor.

[0030] Furthermore, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the multidimensional information event recognition method based on a multi-core support vector machine via a processor.

[0031] Those skilled in the art will understand that the technical solutions provided by this invention may take the form of a method, system, or computer program product. Therefore, this invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this invention may take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, produce an implementation of the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0032] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A multidimensional information event recognition method based on multi-kernel support vector machine, characterized in that, Includes the following steps: S101, extract multi-dimensional feature vectors from the power grid, including some or all of the time-domain feature values ​​and frequency-domain feature values. The time-domain feature values ​​include some or all of the effective values, skewness, and kurtosis of voltage and current. The frequency-domain feature values ​​include some or all of the amplitude spectrum, phase spectrum, and specified h-th harmonic amplitude of voltage and current. S102, input the multidimensional feature vector into the multi-kernel support vector machine to realize multidimensional information event recognition. The kernel function of the multi-kernel support vector machine is obtained by weighted fusion of the kernel functions of Gaussian kernel and quantum kernel. The quantum kernel is the square modulus of the inner product between the corresponding quantum states of different feature vectors in the multidimensional feature vector to quantify the global structural similarity of the feature vector.

2. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 1, characterized in that, The functional expression of the kernel function of the quantum nucleus is: ， in, For the kernel function of the quantum nucleus, and These are multidimensional feature vectors. The i-th eigenvector and the j-th eigenvector in the image. Representing the eigenvector quantum state and quantum state The inner product between For all qubits, the measurement is as follows The probability, For feature vectors The conjugate of quantum feature maps, For feature vectors Quantum feature map, for The initial state of each qubit.

3. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 2, characterized in that, The kernel function of the multi-kernel support vector machine is expressed as follows: ， ， in, For the kernel function of a multi-kernel support vector machine, and These are the kernel functions for Gaussian kernels used to process time-domain features and frequency-domain features, respectively. and These are the kernel functions for quantum kernels that process time-domain and frequency-domain features, respectively. , , , , , It is the weighting coefficient.

4. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 3, characterized in that, The function expression for inputting multidimensional feature vectors into a multi-kernel support vector machine to achieve multidimensional information event recognition is as follows: ， in, This represents the multidimensional information event recognition results from a multi-kernel support vector machine. For symbolic functions, For the weight vector, the superscript... For transpose operation, For the kernel function of a multi-kernel support vector machine, This is a bias term.

5. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 3, characterized in that, The training of the multi-kernel support vector machine includes: S201, extract multidimensional feature vectors and their labels from multidimensional information event samples from the power grid to construct a training dataset; S202, Adjust weighting coefficients , , , , , And construct the kernel function for the multi-kernel support vector machine; S203, train a multi-kernel support vector machine using the training dataset and optimize the model parameters according to the following formula: ， ， ， in, For the weight vector, For bias terms, is the penalty coefficient in the kernel function of the Gaussian kernel. The number of samples in the training dataset, For the first Slack variables for each sample, For the first The label of each sample The kernel function is the kernel function of the multi-kernel support vector machine; the model parameters include the weight coefficients in the kernel function of the multi-kernel support vector machine. , , , , , And the bias term in the kernel function of the Gaussian kernel. and penalty coefficient ; S204, determine whether the preset termination condition has been met. If the preset termination condition has not been met, jump to step S202 to continue training; otherwise, end and exit.

6. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 5, characterized in that, The function expression for the preset termination condition in step S204 is: ， in, For prediction error, , For the new model parameters, For the current model parameters, This represents a multi-kernel support vector machine.

7. The multidimensional information event recognition method based on multi-kernel support vector machine according to claim 1, characterized in that, The kernel function of the Gaussian kernel is expressed as follows: ， in, Let be the kernel function of the Gaussian kernel. and These are multidimensional feature vectors. The first in 3D eigenvectors and the 1st eigenvector 3D feature vectors The width of the Gaussian kernel.

8. A multidimensional information event recognition system based on a multi-kernel support vector machine, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to execute the multidimensional information event recognition method based on a multi-kernel support vector machine as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to execute the multidimensional information event recognition method based on a multi-kernel support vector machine as described in any one of claims 1 to 7 via a processor.

10. A computer program product, comprising a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to execute the multidimensional information event recognition method based on a multi-kernel support vector machine as described in any one of claims 1 to 7 via a processor.

Images