Frequency Analysis on Sound Generated by Automobile Theft and Estimation of Theft Event

Abstract

The authors proposed a security system that could detect two or more theft events ((a) opening a lock, (b) opening a hood, (c) opening a trunk, (d) picking and (e) opening a door) of a vehicle using car speakers. Two or more theft events were distinguished by using the amplitude of the sound. However, this method may be affected by the surrounding environmental sounds. In this paper, the author proposes a new method using sound frequencies to determine theft events. First, the noise in the measurement signal was reduced by wavelet shrinkage. The signal of almost events was able to be detected. To investigate the frequency of each event signal, fast fourier transform was performed. In two or more theft events, there was a difference in the frequency elements. Finally, neural network was used to provide relation between frequency elements and five events. As a result, the method proposed is able to distinguish between the five events. The proposed method was verified experimentally that it is possible to determine some of the theft event.