Facial Expression Recognition Algorithm Research

With the development of artificial intelligence technology, facial expression recognition technique has got much attention.Expression is important carrier of human emotional expression, identification of understanding human expression conveys information will greatly improve the human-machine interaction experience, so the practical face recognition algorithms have very important significance.Aiming at the shortcomings of the existing algorithm, this paper puts forward a new face recognition algorithm, the main innovation points as follows:(1) The algorithms in this paper add the facial feature points localization algorithm expression recognition process, to achieve accurate fast position the face area, and eliminates the algorithm of the entire human face feature extraction methods, only to the facial region feature extraction, reduce the amount of calculation, ascending algorithm in the actual scene detection speed.(2) A new shape feature is put forward in this paper, the interpolation shape features.Lagrange interpolation method is applied to fit the shape of the facial features which are used to represent expression characteristics. Compared with the previous methods, the algorithm is cheaper and faster in calculation.(3) This paper proposes a new texture feature, local gradient intensity direction features.A new kind of compass template was designed to further improve the stability on noise and accuracy of recognition.(4) A new method combines the advantage of shape features and texture features, not only to improve the recognition rate but also robustness to environmental factors such as noise and non mono-illumination.These methods were validated on the extended cohn-kanade dataset, by using the SVM classifier.It turns out they all achieves better recognition effect and higher practicability.