Objectification of dysarthria in Parkinson’s disease using Bayes theorem

This paper presents an assessment of vocal impairment for separating healthy persons from patients with Parkinson’s disease (PD). We have recently shown that deterioration of speech performances in PD speakers is notable from an early stage of the disease, even before starting pharmacotherapy. In this study, we present the potential of the simple Bayes rule to reveal changes in degradable speech performance in the course of PD-related dysarthria. The various speech data were recorded from 23 speakers with recently diagnosed PD and 23 healthy speakers. It has been found that 19 various acoustic measurements are able to differentiate PD significantly from healthy speakers. Subsequently, the Bayes theorem was applied to each of these measurements. As a result, the 21 PD patients and 21 healthy people were correctly classified according to their group. The Bayes theorem thus confirms its feasibility for identifying the features of the impaired voice. Key-Words: Parkinson’s disease, Hypokinetic dysarthria, Speech and voice disorders, Acoustic analysis, Speech signal processing, Bayes theorem.