Research On Facial Expression Recognition Based On Video

With the development of the times,our country has obtained more research results in the field of video surveillance and analysis,and applied to many areas of life.For example,Ping An City intelligence collection and analysis;in public transport,the monitoring system can complete license plate recognition,speed detection and monitoring of vehicles' violation of traffic rules;and so on,in the banking field,monitoring system and its important monitoring anomalies.To ensure the safety of property,these are inseparable from the intelligent video surveillance system.Therefore,for the public monitoring system,we can pay more attention to vulnerable groups,such as the elderly and children.Intelligent detection of abnormal facial expression behavior will be of great significance to enhance national happiness.For video-based facial expression detection,the method in this paper is based on depth learning model,neutral expression of expression and expression component learning.First,the video face is accurately and dynamically tracked by the depth learning training model,then the required facial expression images are obtained,and then the generated model is trained by CGAN to regenerate neutral facial images for any query image.Secondly,the learning program is formed on the inner layer of the generation model,and the learning process can capture the expression components of facial expressions recorded in the generation model.The proposed method is evaluated on different databases and spontaneous expression data sets.Four public facial expression databases were used in our experiment.In our experiments,the method was evaluated on CK + databases.Experimental results demonstrate the superiority of the method.In short,the main contributions of this paper are as follows:1.In this paper,when dealing with dynamic video,the model learning of deep learning is adopted,which can play a prominent role in anomaly detection in video.2.In this paper,facial expression can be decomposed into expression and neutral components.This new perspective is used to verify the new expression recognition method.