Instantaneous Incident Detection System Based on Analysis ofAcoustic Signal from Crash and Skid in Tunnel

Introduction:An acoustic signal-based tunnel accident detection system was developed in this study. In a tunnel environment, the sound diffusioneffect is minimized and thanks to that, discrimination of accident sounds (crash and skid) from other noises can apparently beaccomplished. Discussion:The system is composed of three parts: algorithm, field device, and center system. To distinguish accident-related acoustic signalssuch as a crash or skid among various other sounds in a tunnel, a delicate algorithm that can discriminate those signals from othernormal signals generated from moving vehicles was created. Conclusion:The developed algorithm processes acoustic signals to filter out noises and to identify accident-related signals. The field device,installed in a tunnel, collects analog sounds, transforms them into digital signals, and transmits the digital signals to the server in thetunnel traffic management center. Lastly, in the tunnel traffic management center, the acoustic signal processing algorithm describedabove, installed in a server system, can instantaneously detect accidents. Once confirmed by the system operators, the information onthe detected accidents is intended to be provided to drivers following behind as well as relevant agencies to prevent secondaryaccidents and to respond promptly. The developed system was evaluated in a real tunnel environment using traffic accident soundsacquired from real crash tests. The detection rates were 95, 91, and 80% at distances of 10, 30, and 50 m, respectively with adetection duration less than 1.4 s. Compared to conventional detection systems using loop detectors or video images that have a longdetection time of around 1 min, the developed system can be regarded as superior in that it has an extremely short detection time,which, of course, is one of the most important factors for automatic incident detection systems