Acoustic Accident Detection System

Abstract

In this research we considered the development of a system for automatically detecting and reporting traffic accidents at intersections. A system with these properties would be a great benefit in determining the cause of accidents and could also be useful in determining features of the intersection that relate to safety. A complete system would automatically detect and record traffic conditions associated with accidents such as time of the accident, video of the accident, and the traffic light signal controller parameters. In this project, the basic research required to develop the system has been considered. This involves developing methods for processing acoustic signals and recognizing accident events from the background traffic events. A database consisting of sounds from vehicle crashes, car braking sounds, construction sounds, and traffic sounds was created. We compute the Mel Frequency Cepstral Coefficients as a feature vector for input to the classification system. A neural network is used to classify these features into categories of crash and non-crash events. The classification testing results achieved 99% accuracy.