Feature Extraction and Selection for Real-Time Emotion Recognition in Video Games Players

Different physiological theories and experiments have studied the link between emotions and humans' information provided by biofeedback sensors. However, only few works have been proposed regarding the acquisition of physiological data in order to investigate the emotions of video game players. In this paper, we propose an overview of different features which can be extracted from a set of physiological data acquired from players during video game sessions. Moreover, we provide a method to select only the most important features to use in a generic supervised learning algorithm. With these features, researchers can develop a model able to predict, in real-time, the players' emotions. Thus, we have conducted a set of experiments, in which we have acquired a set of physiological information, and the self-assessed participants' emotional state. On these data, we have applied a feature selection algorithm providing an overview of the most interesting physiological signals and features that should be considered during the studies on emotions in video game research field.