Blind source separation of speech and music signals using harmonic frequency dependent independent vector analysis

Abstract

Proposed is an advanced frequency dependence model for independent vector analysis (IVA) by introducing a harmonic clique model. The proposed model considers the harmonic structure of speech and music signals by assigning strong dependencies among multiples of fundamental frequencies. The proposed model is more effective in separating music and speech signals than both the original IVA and the sub-band local clique IVA.