Improved GA And Pareto Optimization-based Facial Expression Recognition

The human brain has natural capabilities for recognizing face easily,but automatic face recognition is still a very difficult problem for computer.In a real environment,2D face recognition is inevitably influenced by the illumination,pose and expression and these factors become the biggest bottlenecks for the development of the 2D face recognition technologies.Due to the 3D morphable model has been proposed,which makes the 3D face expression recognition popular.How to model a robust and fast 3D face recognition problem is becoming a challenging problem.In order to achieve facial expression recognition with high accuracy,an improved GA and Pareto optimization-based method is used to recognize facial expression.The Haar-like features representation approach and the Bilateral filter are firstly used to preprocess the facial image.Secondly,the uniform LGBP is employed to extract the facial feature so as to reduce the feature dimension.Thirdly,the GA is improved by adjusting fitness evaluation function and a new Pareto objective function is proposed.Based on the improved GA and Pareto optimization approach,the optimal significant features are selected successfully.Fourthly,the random forest classifier is chosen to achieve the feature classification.In 3D facial expression recognition,we map 3D face into two-dimensional space and then use the 2D recognition method to recognize face expression.Subsequently,some comparative experiments are implemented and the experiment results show that the proposed facial expression recognition algorithm outperforms ones in the existing literature in terms of both the actuary and computational time.The contribution of this paper can be summarized as follows:1)The uniform LGBP method is proposed to extract the facial feature so as to reduce the facial feature dimension effectively.2)The GA and Pareto optimization algorithm are combined to select the optimal significant feature.In order to improve the accuracy of the facial expression recognition,the GA is improved by adjusting an appropriate fitness evaluation function and a new Pareto optimization model is proposed,where two objective functions are formulated toindicate the achievements in minimizing within-class variations and in maximizing between-class variations.3)Based on the robust 3D morphable model,a new 3D shape fitting objective function with regularization term is proposed.The new objective function avoids the over fitting problem of the 3D face shape.4)3D face vertices are mapped into the points of 2D space with the help of ISOMAP algorithm.The algorithm can learn the intrinsic geometric structure of the 3D vertices effectively and the 2D face mapped by ISOMAP algorithm eliminates deformation problem.