Research Of Bad Expression Recognition In Distance Education

As technology is rapidly developing, new knowledge is created every time.In order to help people adapt to new environment of modern society. Our nation isbroadcasting distant education to help people to learn in every place and every time.However, space and distance is separated in distant education, teachers can’t get theinformation of students’ emotion. As teachers can’t make the education strategy basedon the emotion of students, it leads to the ideal affection of distant education.This research is aimed at control the affection of distant education for students bygetting the information about the emotion of students. Firstly, this research analyzesthe affection to students by analyzing the mechanism which is produced by emotion. Inorder to get better understanding of relationship between the status of learning and thestatus of emotion, this research analyze these information mentioned. For the status of3-D learning, this research analyzes the characteristics in the emotion of learning andthe facial expression. Among them, three facial expressions is studied in this research.Firstly, by capturing the facial expression of students in learning from camera, thisresearch extracts the characteristics of facial expression and put them into differentclassification for recognizing to detect the bad emotion. This research has made thefollowing the contribution:In the phase of face detection,we gathered students’ negative emotions images in remote education. At the same time,the implementation of image pre-processing to remove illumination effects.Compared with three kinds of face detection algorithm,Thispaper adopts Volia-Jones face detection framework for face detection. The facial positioning and normalization which has detectedextract expression feature based on the detected face region.This study used threemethods: the first method is that use Gabor filters to extract expression features,the filter with5scale8direction.However, the extracted feature dimension is too large,Theref ore, need to be integration of the extracted features.The second method is used LBP toextract expression feature, used the unified model of LBP operator,the radius is2andhave8sampling points.The third method is used HOG to extract expression feature,We use four cell blocks and nine gradient direction to extract feature.In the phase of classification for facial expression, we define the GMKL learningalgorithm to recognize and classify facial expression. In the Cohn-Kanade database ofhuman facial expression, the Euclidean distance and chi-square distance is applied todo the experiments of comparison to get the better result of classification in thisresearch. Using multi-characteristics and single-characteristic to do experiment ofcomparison to prove multi-characteristics can get better result. Finally, we use ourself-build database of facial-expression to do the comparison of classification. Theexperimental result proves that algorithm is practical value...