Emotion Recognition System using Pulse Plethysmograph

Emotion recognition is significant for the advancement and development of modern technology. Existing methods currently in use are less accurate, expensive, and complicated. This system is based on Pulse Plethysmograph (PuPG) signal analysis and brings novelty in the emotion recognition system. In this study, 594 PuPG signals were acquired from various subjects for five emotions under consideration. Data acquisition was performed by attaching the PuPG sensor to the subject's index finger. Raw PuPG signals were preprocessed to remove the noisy components. Then, after comprehensive features analysis, twenty-two features possessing maximum interclass differences were selected to classify emotions under investigation through Ensemble Bagged Trees, K-Nearest Neighbor (KNN), Support Vector Machines (SVM) and Tree classification methods. Ensemble Bagged Trees achieved a maximum average accuracy of90.7% upon extensive experimentation. The proposed methodology is simple, low cost, and accurate as compared to existing methods and provides a new paradigm towards developing emotion recognition systems.