3D Facial Expression Recognition Based On Neural Network

Facial expression recognition has always been a hot topic in the field of pattern recognition and computer vision.At present,researchers have done a lot of research on facial expressions based on two-dimensional images.However,due to the influence of environmental factors such as illumination,posture,expression and shelter on face image acquisition,the recognition rate of facial expression recognition technology still needs to be improved.The emergence and rapid development of the 3D data acquisition equipment has promoted the rapid development of the researchers for the research of 3D data.Facial expression recognition based on 3D data can greatly overcome the problems of face attitude and illumination change in obtaining two-dimensional face images.Therefore,the research of 3D facial expression recognition based on 3D face data and its achievements are increasing.With the study of deep learning,the deep learning has been applied to many fields.However,in traditional recognition tasks such as facial expression,using deep learning instead of manual extraction to describe features,and automatic learning features and classification of features,such research is still relatively few.On another hand,the application of convolution neural network in 3D facial expression recognition is also a generalization of deep learning.Therefore,this paper mainly combines the convolution neural network to study the 3D facial expression recognition.The main work is as follows:First of all,the process of 3D face data pre-processing.In this paper,three-dimensional face data is raster,and an orderly three-dimensional face point cloud data is generated to prepare the facial expression depth map.Secondly,the facial features are extracted.The information of 3D face key points is obtained directly through the depth information.According to the key point information of 3D face,the 3D face point cloud data are normalized and centralization.And then,the 3D face difference components is obtained by subtracting neutral face component from a three-dimensional face point cloud which with a change of expression.Then the depth difference graph of facial expression is generated,and it is used as the feature of 3D facial expression.Finally,the features of facial expression are studied and classified.A new convolution neural network model is designed in this paper.This model is used to test the Bosphorus face database,and 86.66%recognition probability is obtained.In the end,compared with other 3D facial expression recognition algorithms,the algorithm in this paper has a high recognition probability.