Target Detection on the basis of Empirical Wavelet Transform using Seismic Signal

A novel approach is proposed to detect moving ground target using empirical wavelet transform (EWT) as a time-frequency technique. In order to analyze the performance of EWT, the seismic dataset is generated by acquiring the seismic signature of the moving vehicle, i.e., bus. EWT based time-frequency coefficients have been computed from the seismic signals. The number of statistical features has been calculated from EWT based time-frequency coefficients. With the statistical features, bus and noise have been classified using SVM as a classifier. Accuracy, true positive rate, and area under the curve (AUC) have been used as the performance parameters of the algorithm. The AUC of approximately 95%, true positive rate, and accuracy of about 89% have been achieved.