An LSTM based Deep learning model for voice-based detection of Parkinson’s disease

Abstract

Parkinson's disease (PD), a chronic neuropathological disorder, occurs when certain brain cell clusters are not apt to engender dopamine. As a result, people feel difficulty in writing, speaking, walking and performing various other activities. Numerous research investigations have shown that voice impairment is the most underlying symptoms found in the number of Parkinson's disease patients. In this work, we attempt to explore the possibilities of a deep neural network (DNN) and long short-term memory (LSTM) network-based model for predicting Parkinson's disease using a subject’s voice samples. The various simulations were performed on the dataset to exhibit the efficacy of the models along with their comparison to the conventional machine learning techniques. The results obtained show high values of various metrics including an accuracy of 97.12% and 99.03% for DNN and LSTM respectively which strongly suggest their efficiency for the detection of PD.