Game-induced Emotion Analysis using Electroencephalography

Abstract

Over the past few years, machine learning has enabled organizations to develop insights into the psychological aspects of consumer decision-making to enhance their business. In the context of brain-computer interfaces (BCI), one of our research goals is to profile players by invoking emotional responses through video games genres and recording the players’ EEG. It will further enable us to classify emotions over a spectrum representing true states of a user in a decision-making context and capture those emotions to understand specific personality characteristics. Savitzky-Golay filter has been used to clean the non-stationary, non-linear, and extremely noisy signal and recommend it for single-channel EEG devices like Neurosky Mindwave Mobile 2. Sixteen features from time, frequency, time-frequency domains and classified emotions using each domain separately and as a combination called hybrid. SVM, K-NN, and Boosted Trees classifiers have been used where the highest accuracy achieved is 82.26% with Boosted Trees classifier. Our findings propagate that only four emotions (happy, bored, relaxed, stressed) can be classified for a singlechannel EEG device. Two emotions, happy and bored, achieved the highest individual accuracy of 90.01% and 88.76%, respectively. Keywords – EEG, affective computing, machine learning, neuropsychology, video games, emotions recognition, personality-traits.