Analysis of instantaneous amplitude and frequency of EWT modes for automatic target classification

The present article proposes a novel technique to classify moving ground target using the instantaneous amplitude (I.A.) and instantaneous frequency (IF) of empirical wavelet transform modes. Seismic signals of moving ground vehicles are non-stationary. For such signals, the notion of I.A., and IF are of significant practical importance. I.A. and IF provide the time-varying spectral characteristics of a non-stationary signal, which we utilize for ground vehicle detection and classification purpose. We have used empirical wavelet transform as the adaptive wavelet transform based time-frequency analysis technique. The IA and IF features of the EWT modes are computed by the Hilbert transform of the EWT modes. Several statistical features are then extracted using both I.A. and IF parameters that we use in three different classification algorithms. Then we analyze the performance of the proposed technique with the data set provided by the DARPA. Decision trees, random forest, and KNN are used as the classifiers to classify the target as AAV vehicle, D.W. vehicle, or noise. It has been shown that the true positive rate is 99% for noise, 80% for D.W., and 89% for AAV vehicle using random forest classifier.