To Kick or Not to Kick: An Affective Computing Question

摘要

With the Web 2.0 and its services, the users are able to express their own emotions through the production and sharing of multimedia contents. In this context, we can interpret the flood of selfies that everyday invades the web as the will to overcome the limits of the textual communication. Therefore, it becomes particularly interesting the possibility to extract in an automatic way information about the emotional state of the people present in the multimedia content. The knowledge of a person's emotional state gives useful information for the personalization of many online services. Some examples of the areas where such an approach can be adopted are the services for the content recommendation, for the entertainment and for the distance training. In this paper, the techniques of sentiment extraction from multimedia contents will be applied to the sports world. In particular, the aim is to verify the possibility to predict the result of a penalty kick analyzing the player's face. The objective is to investigate the possibility to evaluate in real time the psychic conditions of an athlete during a competition. The system has been tested on an opportunely built dataset and the results are more than satisfying.