Year: 2014
Type: article
Abstract: Duration is one of the important cues for finding the prosodic variation in human speech and other cues are pitch(F 0 ), amplitude(Intensity) and pauses.Two important concept of duration modeling is in play nowadays, 1segmental duration modeling, 2-syllable duration modeling.Since speech is a complex continuous signal, hence finding the boundary of segments and syllables is a manual intensive work and prone to error.As per our observation, it is relatively more error prone to find the segment boundary than the syllable boundary.This paper presents the effect of duration both at the level of syllable and segment and its distinctive role in finding the prosodic variation at the sentence level.The findings could be adopted as a model for implementing the prosodic variation in text to speech synthesis(TTS) and automatic speech recognition(ASR).The approach for finding the the model is purely based on the statistical inference derived from the duration values with respect to other cues extracted from the recorded speech data..
DOI: 10.5120/17015-7296
Authors: Somnath Roy, Nishant Sinha
Institutions: Jawaharlal Nehru University, IMS Ghaziabad
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W2139204175