Automatic Detection of Road Surface States from Tire Noise Using Neural Network Analysis

This report proposes a new processing method for au tomatically detecting the states from the tire nois e of passing vehicles. To detect tire noise, we use a commercially available microphone as an acoustic sensor, which enables us to easily reduce the cost and size in realizing a prac tical detection system. We propose several feature indicators in the frequency and time domains to successfully classify the states into four categories: snowy, slushy, we t, and dry states. The method is based on artificial neural networks. The proposed classification is carried out in multi ple neural networks using learning vector quantization. The outco mes of the networks are then integrated by the voti ng decisionmaking scheme. From experimental results obtained for more than a week in snowy areas, it has been demonstrated that an accuracy of approximately 90% can be attain ed for predicting road surface states.