Analysis of pathological voices by speech processing

Abstract

The authors present a new speech processing software designed for clinical observation of vocal disorders, speech assessments and some neurological pathologies identification. By using a recorded speech sound, we extract the vocal parameters (pitch period, formant frequencies, vocal timbre, jitter, shimmer) which depend on the speaker physical-psychological state and the linguistic structure of language. The obtained results are compared with the normal values in order to an eventual disorder classification, a pathological prediction and prostheses assessments.