An Approach of Automatic Vehicle Classification by Acoustic Wave Based on PCA-RBF

Acoustic wave signals, radiated from moving vehicles, can be used for automatic vehicle classification as an effective source of information. Acoustic wave signals are processed by self-correlation analysis in frequency domain based on Welch spectrum estimation. Original feature vectors of the linear power spectrum are obtained. Principal component analysis (PCA), aiming to reduce data dimension, is utilized to remove the dependencies of original feature vectors and extract main components. With radial basis function (RBF) neural network as the classifier, automatic vehicle classification is realized. Experiments are made on several typical targets, and the results show that the proposed approach is effective.