A computationally cheaper method for blind speech separation based on AuxIVA and incomplete demixing transform
Jakub Janský, Zbyněk Koldovský, Nobutaka Ono · 2016
This paper proposes a modification of an auxiliary-function based algorithm for Independent Vector Analysis (AuxIVA) by applying it to a constrained set of frequency bins where the activity of (speech) signals is high. The de-mixing transform obtained by this approach is incomplete. Its completion is done through the solution of a convex optimization problem known as LASSO. The experiments with blind separation of speech signals show that the proposed method can be twice faster and sometimes even slightly more accurate in terms of signal-to-noise ratio than the original algorithm applied to all frequencies.