Suitability of Articulation Analysis for Extracting Speech Signals Features of Chinese Speaking Patients With Parkinson

Abstract

There is a close relationship between Parkinson's disease (PD) and speech disorders in people with Parkinson's disease (PWP). Most of the previous studies focus on phonation analysis to extract features from speech signals. For Chinese language, though, articulation analysis can capture specific terms that better distinguish PWP from healthy people. In this paper, we put 28 phonation features and 448 articulation features into 10 kinds of classifiers. The results showed that: 1) The articulation features have better performance compared with phonation features; 2) The combination of 40 articulation features selected by LASSO and the Logistic Regression can achieve highest sensitivity at 82.44%.