Appl. Res. Labs. and Dept. of Math., Univ. of Texas, P.O. Box 8029, Austin, TX 78713-8029
This paper will examine in brevity the matching pursuit algorithm proposed by Mallat and Zhang which yields an adaptive signal decomposition that may be used to derive a phase plane representation of a signal. Matching pursuits is closely related to wavelet analysis and may be considered a superset of wavelet multiresolution analysis. A methodology for coherent feature extraction, or denoising, using matching pursuits will be developed and explained. Acoustic signals derived from speech and music will be examined. Synthetic white noise and acoustic noise will be injected and matching pursuits will be used to attempt to remove noise from these signals.