Traffic Flow State Identification Based on Traffic Noise Signals

Abstract:

A vehicle acoustic signal is composed of engine noise, tire noise, air turbulence noise and exhaust noise. Traffic noise signals mixture of multiple vehicles’ acoustic signals is determined by the traffic state. Based on the analysis of the existing traffic state identification methods and vehicle sound characteristics, an identification method of the traffic flow state was proposed by using traffic noise signals. In accordance with the vehicle speed, traffic flow states were divided into the free flow, saturation flow and traffic jams. The spectrums of vehicle acoustic signals were analyzed for different traffic states, the normalized peak frequencies were as features, and different traffic flow states were recognized using support vector machine. The test results show that different traffic flow states can be correctly recognized with traffic noise signals with a high recognition accuracy.