Research On Video Micro-expression Recognition Technology Based On Adaptive Motion Magnification And Deep Learning

As a facial expression that naturally emerges in a very short period of time,micro-expressions contain the true emotions of people.Using microexpressions can effectively help people to identify emotions.Therefore,micro-expressions have a wide range of applications in the fields of justice,business,and consulting..In view of the low efficiency of relying on manual micro-expression recognition,how to efficiently perform automatic microexpression recognition based on computers has become an important research direction.In this direction,due to the extremely short duration and minimal variation of micro-expressions,the recognition performance of existing automatic micro-expression recognition methods still has certain limitations.In view of this,this article will focus on micro-expression recognition technology in further research.First of all,aiming at the characteristics of the small change of microexpression motion,based on the introduction of Learning Based Vedio Motion Magnification(LVMM),the micro-expression video motion amplification technology is improved,and an adaptive-based Microexpression video motion magnification method based on motion magnification factor.Furthermore,combined with this improved method,a video micro-expression expression technology based on adaptive motion amplification is formed.The experimental results show that the technology has a good effect in terms of recognition accuracy.Secondly,to address the limitations of the existing micro-expression recognition technology in the extraction and classification of microexpression features,Convolution Neural Networks(CNN)was used to extract the deep features of micro-expressions,and a restricted Boltzmann machine(Restricted Boltzmann Machine(RBM)and Autoencoder(AE)perform feature dimensionality reduction,and then a micro-expression feature classification technology based on stitching feature maps is given.Furthermore,combining the above-mentioned video motion amplification,feature extraction and classification technologies,a micro-expression recognition technology based on adaptive motion amplification and deep learning is formed.Experiments show that the micro-expression recognition technology based on adaptive motion amplification and deep learning can improve the recognition accuracy.Finally,based on the above research results of micro-expression recognition,a prototype system for micro-expression recognition was designed and implemented,which can be used as an effective tool for microexpression recognition.This system takes micro-expression video or frame sequence as input,and uses the video micro-expression recognition technology based on adaptive motion amplification and deep learning given in this paper to output emotion categories.Experiments show that the system shows good performance in terms of robustness and compatibility.