Speech recognition systems suffer greatly from the presence of additive noise. In this paper a two microphone approach is used as a front end processor to enhance speech corrupted by acoustically added noise signals. Extensive testing of the effect of the algorithm on a continuous speech recognition system was made and the results presented. The results show a reasonable SNR improvement of around 5 dB and an improvement in the recognition rate of the enhanced signal although the performance is not as good as that of the original clean signal. Keywords: Adaptive noise cancelling, continuous speech recognition, Adaptive filtering.
Bibliographic reference. Dabis, H. S. / Wrench, Alan A. (1991): "An evaluation of adaptive noise cancelling for speech recognition", In EUROSPEECH-1991, 1301-1304.