An efficient and robust approach to vehicle classification using wavelet domain seismic signal processing

In this paper, we present a technique for vehicle classification then uses wavelet coefficients of seismic signals. These seismic signals are generated due to vehicle movement on the ground and are obtained by using ground sensors. Our technique models wavelet coefficients as a Laplace random variable and uses its statistics as features for classification. Results show that classification efficiency is higher when wavelet coefficients are modeled as a Laplace random variable, as compared to Gauss ion modeling. The technique has a lesser number of processing steps and lesser-space complexity. A new feature is introduced that captures the changes in amplitudes of wavelet coefficients within a wavelet band.