Signal Extraction from One Microphone Input Using ICA Signal Decomposition

摘要

We propose a novel method for signal extraction from one microphone input using Independent Component Analysis (ICA) decomposition. The proposed method decomposes one microphone input, which is an additively-noise-contaminated signal, into statistically independent components by Finite Impulse Response (FIR) filters, the coefficients of which are determined by ICA. Signal synthesis for the signal part of the one microphone input using the components decomposed then is carried out according to their statistical measures. Simple computer experiments are performed to show the possibility of the method.