Deep Learning Techniques for Parkinson's Detection Using EEG Signals Analysis

Parkinson's Disease (PD) is the prominent reasons of demise in adults over the age of 60. The disease is centered on clinical evaluation. The leading objective of the proposed work is to use Electroencephalogram (EEG) signal analysis for the discovery of Parkinson's Disease at an early stage with the help of Machine learning and Deep learning techniques. The features were extracted by time-frequency representation from the Mel Spectrogram for each of the 32 EEG channels. With experimental analysis, only 12 channels are used, which covers the significant information for PD identification, so to reduce the computational cost, this data is passed into Machine learning and Deep learning algorithms for example support vector machines (SVM), K-Nearest Neighbors (KNN), Random forests, and Convolutional Neural networks (CNN) to determine the Accuracy of the model. A decent accuracy of classification was received, which was in the range of 86-97%. Thus, we propose a cost-effective and risk-free method of detecting Parkinson's Disease with significant accuracy.