A Type-2 Approach in Emotion Recognition and an Extended Type-2 Approach for Emotion Detection

This paper aims at a fuzzy relational approach to similar emotions expressed by different subjects by facial expressions and predefined parameters. Facial attributes represents a wide variety subjected to different circumstances. The fuzzification and mapping of facial features like eye-opening, mouth-opening and length of eye-brow constriction from localized areas is done into emotion space by employing relational models. Uncertainty can be adeptly dealt with fuzzy logic where type-2 approach reigns supreme, which is developed on the basis of various patterns of facial features of various emotions of subjects. The fuzzy free space employs two type-2 fuzzy sets namely interval type-2 fuzzy set (IT2FS) and majority general type-2 fuzzy set (MGT2FS). The former (IT2FS) considers only primary membership functions for m facial features extracted from n subjects which exhibits instantaneous facial expression for a particular emotion. But the latter (MGT2FS) combines the function of IT2FS by introducing primary membership function and implementing secondary memberships for each primary membership curve, thus achieving the desired solution by minimization and optimization. Two other schemes namely average general type-2 approach (AGT-2) and centroidal general type-2 approach (CGT-2) are implied. Experimental results and computer simulation indicates that the proposed theory for emotion recognition and control is robust, easier to implement and possesses higher degree of accuracy.