Abstract

Th is paper describes the concept of automatic segmentation of continuous speech signal. The language used for segmentation is the most widely spoken language i.e. Punjabi. Like all other Indian languages, Punjabi is a syllabic language, thus syllables are selected as the basic unit of segmentation. The traditional way of representing the speech signal is in terms of features derived from short - time Fourier analysis. It is difficult to compute the phase and processing the phase function from the FT phase. By processing the derivative of the FT phase, the information in the short - time FT phase function can be extrac ted. This paper describes the process of automatic segmentation of speech using group delay technique. This includes segmentation of continuous Punjabi speech into syllable like units by using the high resolution properties of group delay. This group delay function is found to be a better representative of the STE function for syllable boundary detection.