Quantification of Prosodic Impairment in Patients with Idiopathic Parkinson's Disease

Abstract

This paper deals with quantitative analysis of prosodic impairment in idiopathic Parkinson’s disease (PD). Experimental dataset consisted of 97 PD patients and 55 healthy speakers. The prosodic features expressing monopitch, monoloudness and speech rate deficits are extracted from stress-modified reading task. Classification accuracies of 70:71% for females, 70:03% for males, and 63:20% for a mixture of both gender were achieved. According to permutation test (1000 permutations, a = 0:01), the models were shown statistically significant. Promising potential of prosodic features to identify HD was confirmed.