Traffic information extraction of vehicle acoustic signal based on neural networks

A method for traffic information extraction of vehicle acoustic signal based on neural networks is proposed. At first, a method of pre-processing and feature extraction of the vehicle acoustic signals is explained, and the Mel-frequency cepstral coefficients are selected as the characteristic parameters of the vehicle signals. Next, the basic theory of the current most widely used neural networks-BP network (Back-Propagation Network) is introduced, and aiming the shortcoming of the BP network, the improvement method to reduce the training time of the network is proposed. At last, the experimental data is used as the sample to train the network, and the target data is recognized. The traffic information is extracted from the target data and the recognized rate can reach 90%.