A fuzzy approach for facial emotion recognition

摘要

Emotions play a critical role in people interaction. The ability to understand human emotions by computer is desirable in several applications, especially by observing facial expressions. This paper presents a new fuzzy-based method for emotion recognition from eyes and mouth features in different ages. The method detects eyes and mouth from combination of different color spaces. Four parameters consist of eye opening, mouth opening, eye opening/width ratio and mouth width are selected for fuzzy analysis. Employing Mamdani-type implication relations, facial attributes and their mapping to emotion space are encoded. Applying the method on Ebner's facial expression database indicates 78.8% accuracy.