Parkinson’s Disease Detection Using Deep Learning

Authors:

Bhimanadham Karthik Reddy¹, Bodam Dharma Teja Reddy², Dr.S.Geetha³, Dr.P.Dhivya⁴, Dr. S. Akhila⁵

¹²³⁴⁵Department of Computer Science,

Dr. M.G.R. Educational and Research Institute, India

Abstract—Parkinson’s disease is a progressive neurological disease that interferes with movement, speech and coordination, so if caught early it may assist in the management and treatment of either condition. This is because Parkinson’s disease produces changes in the pitch, tone and rhythm of the voices of affected individuals so that it is possible to detect the disease through vocal signals. Such subtle changes in voice features prior to the onset of symptoms make it an ideal biomarker for early detection of Parkinson's disease. The voice parameters extracted from sustained phonations in this study included harmonic-to-noise ratio, jitter, shimmer and fundamental frequency. These features pick up subtle variations in how stable and clear the voice is, both known indicators of parkinsonism. The main steps in this project are data pre-processing, features selection, and modelling using a neural network algorithm. The models trained on these recordings can distinguish between a healthy speaker and one afflicted with Parkinson’s, identifying even the most subtle differences in vocal patterns. All-in-all, the results reveal that machine-learning based methods could serve as a highly precise and non invasive complementary tool for Parkinson's disease diagnosis.

Keywords-- Voice Analysis, Data Preprocessing, Neural Networks, Disease Prediction.