Face Features Detection And Face Expression Recognition

Face expression Recognition system is expected to have numerous applications in many fields,In this paper the full system of the Face expression recognition and achieve methodology of the Face expression recognition is discussed in detail.The full system can be divided into some parts: Face expression image capture,Previously processing ,Face feature detecting and locating, Face segmentation and normalization, Face expression feature abstraction,Face expression recognition.In our work,we have done a great deal of experimentation and studying in system of the Face expession recognition except part of Face expression image capture. Main process is: we give a short review on the recent development in the Facial Expression Recognition (FER) research, and then introduce Face features detection,main face organ locating and face segmentation ,Further,we realize Face Expression Recognition with two methods,The first one basic on principle of the Discrete Wavelet Transform (DWT) and the Discrete Cosine Transform (DCT). which is implemented as follows: first, we locate the face, cut it from the whole image and normalize; then reducing the data dimension by DWT, decorrelation by DCT, extracting the feature vector by ZIGZAG; finally, we use the classical Mahalanobis distance to classify expressions. The experiment result shows that: the algorithm in this approach is easy and fast, the maximal recognition ratio for the six basic expressionsreaches 89.44%, The shortcoming of this method is that the generalization is not very good, which still needs improvement. In order to improve efficiency of face expression recognition ,we actualize the second method:It is based on BP Network for Face Expression Recognition,we find efficiency of Face expression is improved greatly and the generalization is enganced.