A Novel Approach for Emotion Classification in Virtual Reality using Heart Rate (HR) and Inter-beat Interval (IBI)

Abstract

Background: The field of emotion research has been progressing rapidly in affective computing and has received much attention from the machine learning community of late. One area that has seen increasing interest relates to the use of physiological data for the prediction of human emotions. In this study, a novel method of hybridizing heart rate (HR) and inter-beat interval (IBI) signals as classification features is presented for classifying emotions into four distinct quadrants in a virtual reality environment. Method: A machine learning approach using a support vector machine (SVM) classifies the sensor data using a combination of HR and IBI data acquired via a wearable device, where the HR and IBI data were combined in a novel manner to realize the classification of emotions in four distinct quadrants. For this experiment, 24 participants participated in the testing where their HR and IBI data were collected while viewing 360° emotional stimuli videos using a Virtual Reality (VR) headset. Findings: The best participant in this experiment achieved an accuracy result of 100% for intra-subject four-quadrant classification while an overall average accuracy of 67.4% was obtained over the entire cohort using this novel HR and IBI signal combination. Conclusion: The findings demonstrate promising results through the use of this novel approach of hybridizing the HR and IBI signals as classification features for predicting emotions in four distinct quadrants with VR as the stimuli. The potential of this research can be applied but not limited to gaming, entertainment, and rehabilitation using VR.