Towards psychologically based personalised modelling of emotions using associative classifiers

Learning environments, among other user-centred applications, are excellent candidates to trial Computational Emotions and their algorithms to enhance user experience and to expand the system usability. However, this was not feasible because of the paucity in affordable consumer technologies that support the requirements of systems with advanced cognitive capabilities. Microsoft Kinect provides an accessible and affordable technology that can enable cognitive features such as facial expressions extraction and emotions detection. However, it comes with its own additional challenges, such as the limited number of extracted Animation Units (AUs). This paper presents a new approach that attempts at finding patterns of interaction between AUs, and between AUs and a given emotion. By doing so, the authors aim to reach a mechanism to generate a dynamically personified set of rules relating AUs and emotions. These rules will implicitly encode a person's individuality in expressing one's emotions.