SINGLE MIXTURE AUDIO SOURCE SEPARATION USING KLD BASED CLUSTERING OF INDEPENDENT BASIS FUNCTIONS
Khademul Islam Molla, Nobuaki Minematsu · 2004
In this paper, we present a technique to separating the audio sources from a single mixture. The system is based on the extraction of independent basis function from the mixture spectrogram and grouping them to produce the source subspaces. Principal component analysis is used for dimension reduction and independent component analysis is employed here to make the basis functions independent from each other. Kullback-Leibler divergence (KLD) based information theoretic clustering algorithm is introduced in this work. The proposed algorithm is suitable for better grouping of the basis functions to separate the individual source. The satisfactory result of two-source mixture separation motivates to use this system for real world single mixture source separation. 1.