Automatic Detection of Parkinson's Disease Based on Modulated Vowels

In this paper we present a novel approach of automatic detection of phonatory and articulatory impairments caused by Parkinson’s disease (PD). Modulated (varying between low and high pitch) and sustained vowels are considered and analysed. The fundamental frequency of the phonations and its range are computed using the Hilbert-Huang transformation. Additionally, a set with “standard” measures are calculated to model phonatory and articulatory deficits exhibited by Parkinson’s patients. Kernel Principal Component Analysis was also applied in order to reduce the dimensionality of the representation space. The automatic discrimination between speakers with PD and healthy controls (HC) is performed using decision trees. According to the results, modulated vowels are suitable to evaluate phonatory and articulatory deficits observed in PD speech.