Preliminary Acoustic Analysis of Noise Components in Patients with Parkinson's Disease

Abstract

This paper deals with acoustic analysis of noise components extracted from speech signals of patients with Parkinson’s disease (PD) who recited a poem. Experimental dataset consisted of 97 PD patients with different disease progress and 55 healthy controls (HC). The analysis is based on parametrization of 2 rhymes recitation using dysphonia features. We obtained classification accuracy 76.66% for female speakers, 69.65% for male speakers and 69.24% for the mixture of both genders.