Sistema multicanal adaptativo para supressão de ruído usando arranjo de microfones.

Abstract

This dissertation consists of a study, evaluation and software implementation of an adaptive noise suppression system, applied to degraded speech signals. An array of four microphones is used, at the system input, to receive the signals. After the signal reception, the signals are phase aligned and then processed. The alignment is made based on the cross-correlation coefficient between the channels, giving a gain of 2 to 4 dB in the signal-to-noise ratio. In addition, the noise suppression is achieved by filtering the signals using an adaptive Wiener-filter. The post-processing of crosscorrelation measurements increases the system performance and an additional gain of about 6 dB at the filter output can be obtained. The enhanced signal, obtained at the filter output, can be used as an input to transmission systems and to speech recognition systems. Informal subjective listening tests indicates an improving in the signal quality and intelligibility.