Phase-space portraits of speech employing mutual information and perceptual masking

Abstract

The use of phase-space portraits for speech has been examined by a number of authors. In this paper we examine the use of speech entropy via mutual information to compute the embedding delay of phonemes and hence construct meaningful phase-space portraits. Since speech signals are known to be spectrally redundant, the effects of perceptual masking on phase-space portraits is also considered. The results indicate that phase-space gives a true indication of the underlying behaviour of phonemes without significant distortion from perceptually masked signal components.