Prediction of Parkinson’s disease from Voice Signals Using Machine Learning

Abstract

Parkinson’s Disease(PD) is a common neurological condition related to the Central Nervous System, that influence the motion of an individual. Normally, Parkinson’s Disease Patients  have low voice volume with monotone quality. To automate the prediction of this neurological condition, audio signals from the UCI dataset repository had been taken. The major features like Harmonic/Noise Ratio, Jitter, Noise/Harmonic Ratio, Shimmer etc were extracted for the study. In the prior work, an accuracy of 83% was obtained by the LSTM  based model on this dataset. To enhance the model accuracy, a combination of  CNN and LSTM were employed in this work. From the proposed study it was analyzed that the combination model was capable exhibited  a better classification accuracy of 85% when compared to the traditional machine learning model like Support Vector Machine and Recurrent Neural Network like LSTM.