Biomedical signals acquisition methodology for emotion classification in virtual reality game environment

Games, as interactive media, afford creators the unique opportunity to respond dynamically to the player’s actions. Envisioning a scenario where the game can discern the player’s emotional state opens up innovative avenues for design. The advent of modern virtual reality (VR) technology introduced a new array of signals that could potentially be harnessed for this purpose. VR systems are characterized by a standardized set of inputs, including a headset, two controllers, and a suite of motion and positional data generated by the user’s interactions within the virtual environment. This spatial data presents a novel opportunity to infer the player’s emotional state by analyzing their movements and postures, which are known to vary with emotional states. The main goal of the work is to develop a methodology for biomedical signals acquisition while playing VR games with special care on emotion induction. The signals are used to classify emotional states. A special VR application has been developed, consisting of four different minigames with distinct mechanics and types. These minigames were carefully designed to represent commonly encountered gameplay elements in VR experiences, ensuring a representative sample of mechanics prevalent in the majority of VR games. The result of the work is a description of the methodology, an application, and a set of signals that were used for emotion classification. The ready-made InceptionTime network model was used for emotion classification. The InceptionTime is a recognized solution for multivariate time series classification. Results prove the usefulness of the methodology and indicate the complexity of the issue of emotion classification in the context of VR games.