Feature Extraction and Facial Expression Recognition Based on Bezier Curve

In recent years, facial expression recognition has drawn more and more attention of artificial intelligence experts and scholars. There are two main methods to character the features of facial expression. The first one is local feature representation method, it uses facial feature points to represent the key facial parts which make marjor contributions to facial expression recognition, such as eyes, eyebrows, mouth, etc. The second one is global feature representation method, it is achieved by modeling the global face. The advantage of the first method is a small amount of calculation and time, but the error due to the feature points tracking and the error caused by the method itself result in lower recognition rate. The second method, in spite of its high recognition rate, takes a long time due to a great amount of useless calculation. This paper proposes a new feature extraction method based on Bezier curve. On the basis of local feature representation and Bezier curves, this method can accurately portray the key parts with few Bezier control-points, and with less point tracking. With much less calculation and more accurate feature, we obtained ideal recognition rate through rigorous experiment.