Segmentation Algorithm Using Temporal Features and Group Delay for Speech Signals

Abstract

Abstract Automatic Speech segmentation is a technique to segment the speech signals automatically into phonemes or syllables which form the basic units of speech. This plays an important role in automatic speech recognition (ASR) systems. In this work, we have implemented a method for automatically segmenting a speech signal using temporal features such as Short Term Energy (STE), zero crossing rate (ZCR) and the group delay of the speech signal computed using the energy for Hindi speech. The syllable boundaries are determined by the group delay algorithm. Analyzing the experimental results for comparing the manual segmentation and automatic segmentation using the group delay, we can infer that the outcome remains similar to the manual segmentation. KeywordsASRSpeech segmentationPhonemesSyllabesSTEZCRGroup delay