Perimeter Security on Detecting Acoustic Signature of Approaching Vehicle Using Nonlinear Neural Computation

Abstract

We propose using a neurobiology-motivated prototype to detect approaching vehicles and to identify the type of vehicles for perimeter security. Acoustic sound of running vehicles is analyzed. Motivated by mammalian auditory cortex studies, the proposed detector processes both spectral and temporal components of input data. With the exact acoustic signature being unknown, a nonlinear Hebbian learning (NHL), one basic and appealing neural learning function found in human brain, is employed for unsupervised learning. This learning rule extracts representative independent features from the spectro-temporal ones and to reduce the dimensionality of the feature space. During learning, synaptic weights between input and output neurons are adaptively learned. Simulation results show that the proposed system can accurately extract critical features from original input data, and can achieve better performance under noisy environments than its counterparts. Additive white Gaussian noise and colored human vowel noise are mixed with vehicle sounds. For any vehicle recognition, the proposed system decreases the error rate to 3% with improvement 21 ~ 34% at signal-to-noise ratio (SNR)= 0 dB, and functions efficiently with error rate 7 ~ 8% at low SNR=-6 dB when its counterparts cannot work properly at this situation. Next for identifying four types of vehicles, the proposed method has workable performance 60 ~ 85% at low SNR= 0 dB and robust performance 80 ~ 100% at SNR= 5,10 dB.