Multilingual Analysis of Speech and Voice Disorders in Patients with Parkinson’s Disease

Parkinson’s disease (PD) is associated with several speech/voice disorders collectively referred to as hypokinetic dysarthria (HD). The main goal of this study is to identify acoustic features that support the diagnosis of PD while being independent of the language of a speaker. We recorded seven speech (e.g. monologue) and voice (e.g. sustained phonation) tasks in a cohort of 59 PD patients and 44 age- and gender-matched healthy controls (HC) speaking Czech or US English. A non-parametric test revealed that the best discrimination power has a measure quantifying the number of interword pauses per minute. In a consequent classification analysis, utilising logistic regression, we observed a drop in the classification accuracy from 72–73% to 67%, when moving from single-language modelling to the multilingual one. The results of this study suggest that especially the prosodic (pause-based) features could play a significant role in the automatic language-independent diagnosis of PD.