Noisy speech segmentation with multiband analysis and recurrent neural fuzzy network

Abstract

This paper addresses the problem of automatic word boundary detection in the presence of variable-level background noise. Commonly used algorithms for word boundary detection always assume that the background noise level is fixed. In fact, the background noise level may vary during the procedure of recording. In order to solve this problem, we propose the RTF-MiFre-based RSONFIN (a recurrent neural fuzzy network) algorithm. Since the RTF and MiFre parameters can extract useful frequency energy and RSONFIN can process the temporal relations, this RTF-MiFre-based RSONFIN algorithm can find the variation of the background noise level and detect correct word boundaries in the presence of variable background noise level. Our experiment results have shown that the RTF-MiFre-based RSONFIN algorithm has good performance in the presence of variable background noise level presence.