Matching pursuit with frame-based psychoacoustic optimized WP-dictionary

This paper presents the method of matching pursuit (MP) with frame based psychoacoustic optimized wavelet packet (WP) dictionary for selecting most relevant components to be used in compact representation transient part of signal. The wavelet dictionary for matching pursuit is composed of functions that are bounded by frame based psychoacoustic adaptive wavelet packet. Psychoacoustic motivated entropy based cost functions allow us to minimize perceptual relevance and adapt wavelet packet structure with reducing dictionary size. The proposed methodology for selecting most relevant components is based on maximising the matching between the auditory excitation scalograms associated with original and modelled signal correspondingly. This technique allows to significant reduce the number of MP atoms in compare with well known techniques based on damped sinusoids and over-complete WP-dictionary.