COMPARISON OF BLIND SOURCE SEPARATION ALGORITHMS FOR CONVOLUTIVE MIXTURES

Samo Begus, Dusan Fefer

Abstract

The goal of blind source separation is to recover independent sources without knowing the originals and the way they are mixed. Three algorithms for convolutive mixtures were tested and compared on anechoic room recordings.<br />
All algorithms separate all possible combinations of signals, except when both signals are noises. The algorithm with natural gradient outperforms other algorithms. The quality of separation increases if an acoustic absorber is placed between the microphones. The mixture of noise and attenuated signal could be successfully separated if two different algorithms are applied one after another.

Keywords

blind source separation, convolutive mixtures, independent component analysis

Download

PWC-2003-TC4-007.pdf

DOI

-

IMEKO TC

TC4 - Measurement of Electrical Quantities

Back to the proceedings