Research On Signal Classification Method Based On BP Neural Network And SVM

In the process of production a large amounts of data signal contains lots of important information, the information related to the safety of the production equipment performance, affect the quality of the product, was the key to detect fault.Once a failure in the process of production, not only can cause great economic losses, will endanger the personal and equipment safety.For the classification of signal processing just can well solve the problem of the safety of the equipment production, classifying data signal, analyzing the useful information in the signal, detecting whether there is a fault in the equipment, to ensure the safety of production, is an extremely important work, is also the main reason to study signal classification.At first, this paper analyses and studies the signal classification method based on BP neural network, BP network design model, combined with PCA dimensionality of wavelet packet denoising processing method, research the improvement of BP classification method based on data preprocessing, at the same time, genetic algorithm and particle swarm optimization algorithm was applied to classification of BP network method, further optimize the BP network model, the improved BP classification method, and through the simulation experiments verify the effectiveness of the improved BP network classification method.For BP network has low accuracy of classification of defects classification method, further analysis is studied based on support vector machine (SVM) classification method, through the reasonable selection of kernel function and related parameters, analyzes the steps of the SVM classification algorithm, and the data preprocessing method is applied to the data processing, key research SVM classification method based on data preprocessing, and the genetic algorithm and particle swarm optimization (pso) combined with SVM classification method respectively, further optimize the combination of the SVM parameters, improve the accuracy of the SVM classification method.Through the experimental simulation verify the feasibility and effectiveness of the improved SVM classification method.Finally, respectively from the run time and to improve the classification accuracy of two standard of the classification of the BP network method and the improved SVM classification method on the basis of the comparison and evaluation, the simulation experiment results show that the SVM classification method is more superior to BP network method.