Research On Expression Feature Extraction And Expression Feature Selection

With the rapid development of computer science and information technology in recent years,the research of pattern recognition and artificial intelligence has become the academic focus.How to extract and classify facial features accurately and effectively has become a research hotspot.Because of the particularity of human face and the variety and complexity of expression,the extraction of facial features is a key part in the process of expression recognition.Therefore,this paper mainly studies how to extract facial expression features efficiently,designs and implements a real-time remote expression monitoring system at the same time.The main content of this paper is composed of the following three parts:Firstly,propose a facial expression feature extraction algorithm: DCLM.The DCLM feature refers to the difference between the shape of facial organs under the expression and that under the neutral expression,which can be divided into two types: direct DCLM feature and indirect DCLM feature.The direct DCLM features directly utilize the change of feature points,while the indirect DCLM features utilize the difference of distance caused by the change of expression between some feature points.On the premise that CLM can be used to locate key feature points accurately,feature can capture subtle expression changes and is easy to compute.Secondly,propose a facial expression feature selection algorithm: SVMRFE.On the basis of extracting DCLM features,feature selection is made on the basis of feature selection.According to the recursive elimination rule,feature data sets are selected and the expression feature subset is obtained.The experimental results show that the redundant features can be eliminated effectively and the feature containing high information content can be preserved,and the accuracy rate of recognition can be higher.Thirdly,design and implement a real-time remote expression monitoring system.Based on DCLM and SVMRFE algorithm,a real-time remote expression monitoring system based on object-oriented design idea is designed and implemented.The system is based on C/ Sarchitecture.Emoticon data is transmitted to the server through the monitoring device.The server saves the emoticon data and processes the data.At the same time,the facial expression data and its processing results are transmitted to the monitoring platform.Users can monitor the facial expression in real-time and remote way in this way.