Machine learning-based study of eye features under the emotion of anger

Based on the theory of "the liver is open to the eyes", this study uses a machine learning approach to explore methods and models that can be applied to the identification of eye features for angry emotions, in the hope of providing some ideas for objective identification of angry emotions. We used the angry emotion faces in the China Affective Picture System (CAPS) as the research object. Using multiple feature point inspection, we use openface software to obtain the 3D coordinate point coordinate information of the eye features of the angry emotion images. The aggregated data were then statistically analyzed using matlab software to calculate the eye feature values to be used, and then continued to process them using the classifier function to build a face eye feature point model based on Support Vector Machine (SVM), Linear Discriminant Analysis(LDA), K-Neares Neighbor, Ensemble Subspace Discriminant (ESD), Decision Tree(DT) and other classifiers, and the model was comprehensively evaluated in terms of accuracy, area under the ROC curve and other evaluation metrics. For the results of the measured classification tests, the correct rate obtained by DT classification reached 83.1%, KNN reached 84.7%, ESD reached 88.1%, SVM reached 84.7% correct rate and LDA reached 88.1%. The effect of the eye feature description parameters on the classification effect of anger emotion was verified through experiments, and the best performance of ESD and LDA classification was achieved. Based on the above results we can draw the following conclusions. "The eye is the heart's ambassador", and the classifier constructed based on the eye diagnosis information can play the role of classification, which has certain scientific value and can provide objective diagnostic guidance for clinically relevant psychosomatic problems.