Selective noise cancellation using independent component analysis

Abstract

We propose a new ANC system that selectively cancels only the noise signal in the mixture at a specific local position. The BSS separates the desired sound signal from the unwanted noise signal and is used as a preprocessor of the proposed ANC system. In order to enhance the performance of noise separation, we propose a teacher-forced BSS learning algorithm. The teacher signal is obtained form a loudspeaker of the ANC system. Computer simulation and experimental results show that the proposed ANC system effectively cancels only the noise signals from the mixtures with human voice. © Springer-Verlag Berlin Heidelberg 2003.