Recently, all aspects of human computer interaction in the multimedia frameworks, try to be adapted with all feedbacks that can be detected from users. This adaptation can take the name of a social dimension. Amongst these feedbacks, the most important and the most credible we find the emotional back. In that topic this paper deals with the problem of Multimedia Emotion Recognition System (MERS). Based on real time analysis, of non verbal upper body expressive gestures; our system try to compute a probability of the emotional state of a user using simple multimedia peripherals.