Kumar, Sunil2017-12-112023-10-202017-12-112023-10-202017ROLL NO.126102003https://gyan.iitg.ac.in/handle/123456789/887Supervisor: M.K. BhuyanRecognition of human’s emotion through facial expressions has many important applications ranging from behaviour recognition, human-computer interaction, security, psychology, and so on. Recognition of facial expressions from non-frontal faces, and recognition from different views are two important research challenges. As different views of a facial expression are just different manifestations of the expression, the information embedded in different views can be effectively utilized for facial expression recognition (FER). Motivated from the above mentioned facts, we proposed to extract facial informative regions and discriminative shared space for facial expression recognition.Extraction of discriminative features for different facial expressions is a key step in facial expression recognition. However, most discriminative facial features can be extracted from the informative regions of a face. In this view, the importance of different facial sub-regions is investigated, and subsequently the facial sub-regions which have significant contributions in different facial expressions are only considered for feature extraction. Furthermore, a weighted-projection based local binary pattern (WPLBP) feature is proposed. For this, texture features are extracted from informative regions and they are weighted on the basis of their importance. Finally, an efficient face model is derived from the informative regions of a face. The proposed face model has several advantages, and it gives better performance than other existing face models.enELECTRONICS AND ELECTRICAL ENGINEERINGThesis