Analysis of Vocal Tract Shape Variability for Different Conditions in Indian English Vowels

摘要

The study presents vocal tract shape estimation for vowels. Autoregressive model of vocal tract has been used herein for different recording conditions. Vocal tract shapes are known for phonetic distinction and speaker exclusiveness among vowels found by individual speakers ingrained in them. The researchers propose to use this autoregressive model for utterances made by speakers from south India taken on different occasions and in various recording conditions. Samples of speeches have been recorded assuming ice cold water and having an instant check of its effect and again later after 5 min. Vocal tract shapes so obtained are compared to those recorded under normal conditions. Vowel utterances made by speakers have been recorded 30 times, variations in the resulting vocal tract for different conditions on intra speaker basis. The researchers propose to investigate those and also, the variability as a parameter for the speech recognition process. The entire process has been done using MATLAB.