Research On Micro-expression Recognition Algorithm

As an important field of human-computer interaction,facial expression recognition has been developed for several decades and has been widely used in many fields.In recent years,people begin to study a kind of face expression which called micro-expression.Micro-expression is a kind of special expression,which has a short duration,weak intensity,and reflects the real emotion of a person.It is widely used in the field of lie detection,clinical diagnosis and interrogation.The related research of Micro-expression is carried in this paper,by using the static images,dynamic sequence and deep learning.(1)The micro-expression is pretreated.The database used in this paper is CASME2 and SMIC micro-expression data,the scale of the micro-expression image in the database is normalized and gray normalization.(2)Micro-expression recognition based on static image is studied.The maximum expression of each sample in the database is selected as the static micro-expression of the sample,and then LBP and LPQ features are extracted and fused.The experimental results show that the fused features have a great improvement on the recognition rate of the micro-expression.(3)Micro-expression recognition based on dynamic sequence is studied.The LBP_TOP operator is used to extract the expression features of the dynamic sequence,and the LLE algorithm is used to reduce the dimensionality of the high-dimensional feature.The LBP_TOP operator can extract the information of the micro-expression on the time dimension,and the recognition rate is higher than that of the static image.(4)Micro-expression recognition based on depth learning is studied.The micro-expression sequence is inputtd in the 3D-CNN network to extract micro-expression features,and finally use the SVM classification.Compared with other depth learning methods,3D-CNN can directly deal with video or image sequences,the calculation is simple and efficient.