Abstract. Affective and human-centered computing have attracted a lot of attention during the past years, mainly due to the abundance of devices and environments able to exploit multimodal input from the part of the users and adapt their functionality to their preferences or individual habits. In the quest to receive feedback from the users in an unobtrusive manner, the combination of facial and hand gestures with prosody information allows us to infer the users' emotional state, relying on the best performing modality in cases where one modality suffers from noise or bad sensing conditions. In this paper, we describe a multi-cue, dynamic approach to detect emotion in naturalistic video sequences. Contrary to strictly controlled recording conditions of audiovisual material, the proposed approach focuses on sequences taken from nearly real world situations. Recognition is performed via a 'Simple Recurrent Network' which lends itself well to modeling dynamic events in both user's facial expressions and speech. Moreover this approach differs from existing work in that it models user expressivity using a dimensional representation of activation and valence, instead of detecting discrete 'universal emotions', which are scarce in everyday human-machine interaction. The algorithm is deployed on an audiovisual database which was recorded simulating human-human discourse and, therefore, contains less extreme expressivity and subtle variations of a number of emotion labels.