Non-invasive player experience estimation from body motion and game context

In this paper, we investigate on the relationship between player experience and body movements in a non-physical 3D computer game. During an experiment, the participants played a series of short game sessions and rated their experience while their body movements were tracked using a depth camera. The data collected was analysed and a neural network was trained to find the mapping between player body movements, player in- game behaviour and player experience. The results reveal that some aspects of player experience, such as anxiety or challenge, can be detected with high accuracy (up to 81%). Moreover, taking into account the playing context, the accuracy can be raised up to 86%. Following such a multi-modal approach, it is possible to estimate the player experience in a non-invasive fashion during the game and, based on this information, the game content could be adapted accordingly.