Research On Micro-Expression Recognition Based On Convolutional Neural Network

With the rapid development of artificial intelligence,a series of biological information recognition technologies such as face recognition and fingerprint recognition have been applied in people's daily lives,They bring convenience to people's lives.Facial micro-expression recognition as an important and difficult technology in facial recognition technology has attracted more and more researchers' attention.Micro-expressions can reflect a person's mental state more truly due to its own irrepressible and uncontrollable characteristics.Therefore,it has great application potential in the fields of polygraph,criminal investigation and psychotherapy.However,the short duration of the micro-expression and the small amplitude of the expression action make the traditional image recognition algorithm perform unsatisfactory on the task of micro-expression recognition.The good performance of convolutional neural network in the field of image processing lets people begin to apply it to the task of micro-expression recognition.This paper mainly studies the micro-expression recognition algorithm based on convolutional neural network,and proposes solutions to problems such as the unobvious expression of micro-expression actions,the difficulty of convolutional neural network to extract video dynamic features and the redundancy of micro-expression video information.The main work of this paper is as follows:(1)A micro-expression recognition based on video amplification algorithm and optical flow network is proposed.First,video amplification algorithm based on deep learning is used to amplify the motion range of the micro-expression,and then the optical flow image of the video is extracted by the optical flow network Lite Flow Net as the video dynamic input network,which solves the problem that the convolutional neural network is not suitable for extracting the dynamic features of the video.Send it to the designed convolutional neural network for recognition.(2)A micro-expression recognition based on improved dual-stream shallow convolutional neural network is proposed.The network adopts dual-stream input,which can input the magnified gray-scale image and the optical flow image at the same time.The network merge two features to recognize micro-expression and improve the recognition accuracy.Atrous convolution and attention module are also introduced into the network to improve the ability of the network to extract high-dimensional effective features.(3)In order to improve the acquisition of effective information,the initial frame and key frame of the video are used to replace the entire video during the experiment,which reduces the impact of redundant information in the video on recognition and improves the recognition efficiency.At the same time,the data is amplified to meet the data requirements of convolutional neural network training.In this paper,three databases of CASME?,SMIC-HS and SAMM are used for experimental verification,and the results show that the algorithm in this paper can effectively improve the accuracy of micro-expression recognition.