Bezier curve and decision tree modelling for facial expression analysis

It is crucial to handle missing or incomplete information obtained from various low level image processing tasks. Completion of such information requires foundation of typical mathematical, geometrical and linear algebra concepts. In this paper, emphasize is on completion of missing information in terms of detected pixels for expression recognition. Amongst seven standard expressions we chose to work with only Happy, Sad and Normal in this paper. Feature extraction, feature completion and trained decision tree model in XML format on novel and extended dataset has been demonstrated though classical edge detection and Bezier curve modeling. Experimental results show that the recognition rate of proposed system is 71% on CK data set, 76% on JAFEE dataset and 78% using PAKFE dataset for chosen principal emotions.