Evaluation of LPC trajectory for Vowel-Consoant-Vowel sequence

Abstract

This paper shows an analysis of Linear Predictive Coding (LPC) coefficient for Vowel-Consonant-Vowel (VCV) sequence in order to determine the temporal transfer function of vocal tract. The Linear Prediction model is a linear combination of the speech sample and the glottal pulse as exciter. The acoustical output of the dynamic system is speech sample. Here, a sequence VCV is analyzed by using LPC coefficient and their mean, standard deviation and skewness. The investigation, focused on the variation of LPC coefficient can be applied successfully for speech recognition and vocal tract dynamics modeling.