A comparison of molecular approaches for generating sparse and structured multiresolution representations of audio and music signals

Abstract

We compare the characteristics and performance of joint (single-step) and sequential (two-step) approaches for creating sparse and structured acoustic signal representations derived using overcomplete methods (OMs). A joint approach, such as molecular matching pursuit (MMP), attempts to find coherent structures in a signal as part of the decomposition process, while a sequential approach, such as agglomerative clustering (AC), attempts to find coherent structures after the signal decomposition. We review each approach, and examine their performance using real audio and music signals.