Hierarchically structured model and basic algorithms of time domain segmentation of speech signals

This paper introduces a systematic approach to speech signal (SS) segmentation based on developed hierarchically structured SS-model. Ground is given to application of multilevel time voice sound record segmentation, and basic algorithms of principal segmentation stages are developed: VAD algorithm; “voiced / noise / plosive” segmentation; voiced fragments’ segmentation into separate pitch periods. The possible variant of splitting adjacent voiced sounds using analysis of their structure has been proposed.