Research On Transfer Learning Model From Macro-Expression To Micro-Expression

With the development of the image recognition technology in the information society,more and more attention has been drawn to micro-expression.As a tiny expression with a very short duration,it is not controlled by mind and differs in many ways from the micro-expressions.Because of its own features,micro-expression is playing an increasingly important role in the field of national security,social stability and judicial interrogation.Although micro-expressions have grown more rapidly in recent years,it is not easy to capture micro-expressions because of their features.The collection of micro-expressions is more complicated and the artificial coding cost is higher,which will lead to the lack of a good micro-expressions database.The scarcity of samples also limits the development of micro-expression recognition.With the development of the research,there is a higher and higher demand for micro-expressions database,people are eager to establish a high quality and high standard micro-expressions database,hoping to propose a new method of micro-expressions recognition,to recognize the micro-expression more efficiently.In view of the above problems,this paper analyzes the features of the micro-expression database,summarizes the achievements and experiences of the predecessors,and synthesizes different test methods.The SDU micro-expression database has the largest number of samples and high image resolutions.At the same time,in order to improve the micro-expressions recognition effectively,this paper also designs the transfer learning model from macro-expression to micro-expression,the main work and innovations in this paper include the following aspects:First,we have established a micro-expression database.The micro-expression database has a total of 300 samples,providing a rich sample of micro-expressions.Second,in terms of the general lack of micro-expressions samples and incomplete label information,this paper presents a transfer learning model from macro-expression to micro-expression,and introduces the idea of multi-view and multi-task,it is hoped that a reliable model can be established by using rich macro-expression information,and the knowledge theory is transferred to the recognition of micro-expressions.The experimental results are given in this paper,it fully proves the reliability of the algorithm and the improvement of the database...