Automatic Facial Expression Recognition Based on Hybrid Approach

The topic of automatic recognition of facial expressions deduce a lot of researchers in the late last century and has increased a great interest in the past few years. Several techniques have emerged in order to improve the efficiency of the recognition by addressing problems in face detection and extraction features in recognizing expressions. This paper has proposed automatic system for facial expression recognition which consists of hybrid approach in feature extraction phase which represent a combination between holistic and analytic approaches by extract 307 facial expression features (19 features by geometric, 288 feature by appearance). Expressions recognition is performed by using radial basis function (RBF) based on artificial neural network to recognize the six basic emotions (anger, fear, disgust, happiness, surprise, sadness) in addition to the natural. The system achieved recognition rate 97.08% when applying on person-dependent database and 93.98% when applying on person-independent.