Volume 06,Issue 03

Parkinson's Disease Detection Based on Signal Processing Algorithms and Machine Learning

Authors

Atiqur Rahman , Aurangzeb Khan, Arsalan Ali Raza


Abstract
There is more interest in the speech model applications for analysis of Parkinson’s diseases to predictively construct tele diagnosis and telemonitoring models. This motivated us to use a relatively large database of voice samples having different types of vowel phonations, compiled from a series of verbal trials for people suffering from Parkinson’s disease. Two main problems are observed in learning from a dataset of this type that contains multiple discourse registration by subject: a) How to foresee such different kinds vowel samples in the diagnosis of Parkinson’s disease (PD)? b) How aptly the main inclination and dispersal the metrics can be used as representatives of all example recordings of a subject? This article examines the multiple types of vowel samples collected from PD patients and healthy subjects and utlize state-of-the-art signal processing algorithms like Perceptual Linear Prediction (PLP) and ReAlitive SpecTrAl PLP (RASTA-PLP) for feature extraction purposes. The extracted set of features are classified using SVM model with four different types of kernels. Results show that our algorithm performs 74% accurately.

Keyword: Parkinson’s disease, Feature Extraction, Machine Learning, Voice Data, Tele diagnosis, Tele-monitoring.

PDF [ 330.6 Kb ] | Endnote File | XML

CRPASE: TRANSACTIONS of



Follow Us

Google Scholar   Academia

JOURNAL IMPRINT