Facial expression classification with Haar features, geometric features and Cubic Bézier curves

Facial expressions are nonverbal communication channels to interact with other people. Computer recognition of human emotions based on facial expression is an interesting and difficult problem. In this study, images were analyzed based on facial expressions and tried to identify different emotions, such as smile, surprise, sadness, fear, disgust, anger and neutral. In practice, it was used Viola-Jones face detector used AdaBoost algorithm for finding the location of the face. Haar filters were used in finding the eyes and mouth. In cases where erroneous detection of the mouth and eyes, facial geometric ratios were used. Cubic Bézier curves were used in determining emotion. FEEDTUM facial expression database were used for training and testing. The seven different emotions used for the study, the recognition success rates ranged from 97% to 60%