Noise Reduction in Reverberant Environments with a Blind Source Separation Algorithm

In this paper we propose the use of a blind source separation (BSS) algorithm as a means to enhance convolutively mixed audio signals by separating the interference signals, such as cars, footsteps, trains, etc., from the desired speech source. Computer simulations confirm that the algorithm adopted in this work can efficiently extract the desired speech signal from a convolutive mixture. These results encourage its possible use as a speech enhancement front-end for automatic speech recognition (ASR) systems for example.