Automatic estimation of severity of Parkinson's disease based on speech rhythm related features

Diseases, such as Parkinson, impairs cognitive processes of patients, through which speech is also affected. In this paper, we propose a method for Parkinson's disease severity level estimation based on speech rhythm related features extracted from running speech (read texts and monologue) uttered by Hungarian Parkinson patients and healthy control population. Classification and regression models are built using various machine-learning methods for both linguistic types separately. Separate and joint decisions were made for the different text types. The final prediction was obtained by fusing the separate estimations for each speaker. Test trials were run in order to investigate, if age is a relevant feature for the machine learning tasks. It was found that the investigated features are useful and highly relevant for the automatic diagnosis of Parkinson's disease based on the classification and regression performances. The best results were obtained using support vector machine (and regression) with 84.62% accuracy for binary classification and 0.735 Spearman correlation for Parkinson severity level estimation measured on the Hoehn-Yahr scale.