|Doctoral student, email: imed.LAARIDH(at)univ-amu.fr|
|Detection of abnormal zones in pathological speech: contribution of the automatic speech processing against the Human expertise|
This project aims to study how automatic speech processing tools could be applied in speech disorder context. Particularly, this work addresses the limits of the current perceptual evaluation methods for dysarthric speech. Indeed, such methods are found to be subjective and unreproducible, thus this work tries to provide objectives measures for the evaluation of the speech quality.
Perceptual evaluation methods are also highly time-consuming which make them unadapted in large corpora and research context. This project aims at providing phoneticians and clinicians with automatic speech processing tools in order to help them in their manual investigation and clinical practice by focusing their attention on speech zones and segments (phonemes, syllables, words) regarded as abnormal or presenting abnormal acoustic patterns.
Given the variability of phenomena observed on dysarthric speech and the variety of pathology and/or dysarthria classes, the efficiency of the automatic approach in locating abnormal zones compared to human experts along with the pertinence of the highlighted segments for phoneticians/clinicians are also a subject of interest of this work.