Voiced/ unvoiced speech discrimination using symbolic dynamics

The aim of this study is to evaluate how far the nonlinear symbolic dynamics approach helps to characterize the nonlinear properties of speech and thereby discriminate between voiced and unvoiced speech segments. The symbolic dynamics calculations were performed on voiced speech, unvoiced speech and silence data. Differences were found in histogram properties and complexity measures of symbol sequences among the three groups. The results of the analysis suggest that the nonlinear symbolic dynamics approach is helpful in classification of speech segments.