Inference Model of Facial Expressions and Emotions Considering Situations

Abstract In human communication, non–verbal information, especially, facial expressions are indispensable since facial expressions reflect emotions well. Even in the communication between human and computer, non–verbal information should be important factor. This paper considers an inference model of human emotions from not only facial expressions but also the situation in which human is put. Neural network models are employed as the inference model. The questionnaire, in which subjects are asked to evaluate emotions given faces by line drawing and the situation, is performed. Using data by the questionnaire, inference models are obtained. In order to confirm the performance of the models, two kinds of questionnaires are done. One is the questionnaire about the evaluation of model results. The other is the questionnaire about the comparison of model inference results and subjects results. The analysis results show that the models can infer emotions well and that the inference model from facial expressions...