Voice analysis for detecting persons with Parkinson’s disease using MFCC and VQ

Abstract

© 2005 - 2014 JATIT & LLS. All rights reserved.In order to improve the assessment of speech disorders in the context of Parkinson's disease, we have used 34 voice recordings of sustained vowel /a /, from 34 subjects including 17 patients with Parkinson’s disease. We subsequently extracted from 1 to 20 coefficients of the Perceptual linear prediction (PLP) from each subject. The frames of the PLP were compressed using vector quantization, with six Codebook sizes. We used Leave One Subject Out validation scheme known as (LOSO) and the Support Vector Machines (SVM) classifier with its different types of kernels, (i.e.; RBF, Linear). After viewing the variety of obtained results, we proceeded to a bench of 100 trials. The best average result obtained was 75.79%, and the maximum result obtained was 91.17% using the codebook size of 1.