Abstract
This paper focuses on the architecture of affective multimodal user interfaces that exist in e-learning applications. The OO (Object Oriented) approach has been adopted in order to combine evidence from multiple modalities of interaction and data from emotion stereotypes and classify them into well structured objects with their own properties and methods. The proposed architecture can be adopted in future multi-purpose emotion recognition systems with multiple modalities and improved emotion detection algorithms. Furthermore, the resulting emotion detection server is capable of using and handling transmitted information from different sources of data during human-computer interaction. The evaluation of the OO architecture for affective interaction by programmers and teachers revealed that there are considerable improvements in the resulting e-learning system, including easiness in adding a modality of interaction, better structure for the available multi-modal information and user friendlier environment for the human-computer interaction. The evaluation of the resulting system provided quite satisfactory results considering the proposed affective multimodal architecture.