Facial Emotion Classification of Multi-Type Datasets based on SVM Classifier
Amornvit Vatcharaphrueksadee, Maleerat Maliyaem, Rattikan Viboonpanich, Suphot Phuangkamnerd · 2022
This paper presents an approach to using 2D images of stylized characters with annotated facial expressions and human facial expressions from the Facial Expression Research Group 2D Database (FERG-DB) and Extended Cohn-Kanade Database (CK+) consecutively. This is to improve the quality of Facial Expression Recognition (FER) for classifying facial expressions of human faces and cartoon animated faces, which will benefit the improvement of acting towards the correct emotion. A geometric and appearance hybrid approach to feature extraction was used to gain essential components. Principal component analysis (PCA) was used to reduce the dimensionality of the features. Support Vector Machine (SVM) classified selected features into six emotions: fear, anger, joy, sadness, surprise, and contempt. The system achieves a recognition rate of 92.63%.