The Research of Vehicle Classification Using SVM and KNN in a Ramp

There is an important significance of the application for real-time classification by using of the acoustic and seismic signals generated by vehicles in the road ramp. The eight test points were put on the both sides of a road ramp, the some devices of acoustic and seismic sensors etc were put in each point. On the acquisition of acoustic and seismic signals, short-time Fourier transform (STFT) was used for feature extraction. In the classification, radial basis function (RBF) kernel was used to train SVM, KNN and SVM were used for the comparative study of real-time classification and achieved good results. We also proposed an improved SVM algorithm which has improved the classification accuracy of SVM to nearly 1 percent. This paper also discussed the classification of the different window size, and discussed the influence on the classification accuracy changes in the window size. And it finally comes to the conclusion by experiment: it is obvious that the classification accuracy is sensitive to the window size. In the classification accuracies, the performance of SVM is superior to that of KNN, and improved SVM is slightly superior to SVM.