Abstract: Seismic signals generated by ground target movement can be used for target location and recognition. When there exist multiple types of ground targets, the signal aliasing phenomenon may interfere signal analysis. EWT algorithm divides the spectrum into several segments according to its maxima value, and transform the signals by constructing orthogonal wavelet functions in each segment to obtain multiple empirical mode signals. However, if the segments are merely devided according to the maxima value, dense peaks may result in too many modes. To solve this problem, spectrum envelope is extracted repeatedly to reduce the number of peaks, and then spectral threshold segmentation algorithm is proposed to enhance discrimination among different signals. Experiments show that improved EWT algorithm has better separation effect for seismic signals from different targets.

Key words: seismic signal, empirical wavelet transform, envelope spectrum, signal separation