Towards a robust real-time emotion detection system for intelligent buildings

The last few years have witnessed an increasing interest from computer scientists in the role emotions could play in the adaptability of artificially intelligent mechanisms. Evidence from neurologists suggests that affective states are crucial in the interaction between an individual and the environment. Furthermore, emotions often dominate our actions and some times override reasoning in the process of making decisions. In this paper an analysis of the use of Autoassociative Neural Networks (AANN) in the context of real-time physiological emotion detection for intelligent inhabited environments (IIE) is presented. Two main studies were undertaken: On one hand the effects of altered physiological responses stemmed from various degrees of emotive intensity and on the other hand the possible consequences of physical arousal not related to emotional expressiveness It is argued that the use of AANN contributes to an improved separation of emotional classes and a more accurate recognition of affective states in individuals with a varying degree of emotional responsiveness. It is also postulated that AANNs robustness is not affected by physiological disturbances associated with physical activities thus setting the basis for emotion recognition in real-life scenarios.