Recurrence Quantification Analysis based Emotion Detection in Parkinson's disease using EEG Signals

Emotional disturbances are Parkinson's disease (PD) patients is typical, and this work aims to identify the emotional disturbances in PD using Electroencephalogram (EEG) signals. Clinicians assess the emotional impairment in PD using International standard questionnaires, and most of the time, this assessment becomes inaccurate since the verbal responses of PDs are not precise to express their internal feelings. EEG based emotional impairment detection in PD gained significant attention due to its robustness, flexibility, and non-invasiveness. In this work, we utilized the EEG dataset consists of 20 subjects each in PD and 20 Normal Control (NC), and EEG signals are collected using 14 channel wireless EEG device over six types of emotions (happiness, sadness, anger, fear, disgust, and surprise) at a sampling rate of 128 Hz. The 6th order IIR Butterworth filter with a cut-off frequency of 0.5 Hz-49 Hz is used to filter the noises and other external interferences. Two features from Recurrent Plot (RP) such as, Maximum Diagonal Line Length (MDLL) and Maximum Vertical Line Length (MVLL) are extracted from alpha, beta, and gamma frequency bands of EEGs. These emotional relevant features are mapped into corresponding emotions of PD and NC using the Probabilistic Neural Network (PNN) classifier. The gamma frequency band (30-49 Hz) feature of maximum diagonal line length gives a maximum mean accuracy of 91.38% and 87.55%, for NC, and PD subjects, respectively.