The Matching Pursuit Method for Extracting Feature Based on DT-CWT and its Application

Abstract

A matching pursuit method based on Dual-Tree Complex Wavelet Transform (DT-CWT) is proposed for extracting feature. Many new orthogonal wavelet bases formed Hilbert transform pairs is constructed by the method which is based on the sufficient and necessary condition on constructing wavelet, via the flat delay filter, and translated the problem into resolving algebraic equations. And taking these wavelets as choice object, a matching pursuit method based on DT-CWT is used for extracting feature. The matching pursuit method is based on series expansion of the signal by a set of elementary functions of orthogonal wavelets formed Hilbert transform pairs to match feature more effectively. Simulation testing and field experiments confirm that the proposed method is effective especially in extracting impulsive feature on high intensity noise, which matching pursuit method based on Discrete Wavelet Transform and other wavelet de-noising methods based on threshold and frequency-band, etc cannot do it completely.