Application of Poincare Map-Based Description of Vowel Pronunciation Variability for Emotion Assessment in Speech Signal

Abstract

An objective of the following paper is to examine feasibility of using vowel pronunciation variability as a descriptor of an emotional state of a speaker. Variability is assessed using Poincare Maps, constructed for pitch-synchronized speech segments that correspond to the vowel "a ". Recognition phase of the proposed approach involves an application of Hidden Markov Models, which are applied to an analysis of temporally-ordered Poincare Map entries.