Research on Facial Expression Recognition Based on Active Learning
Yan Wan, Hongjie Ni, Xiaodong Peng, Yining Guo, Li Yao, Xiangyang Feng · Journal of Physics Conference Series · 2018
In order to solve the problems of low recognition accuracy and poor robustness in traditional facial expression recognition research methods, this paper presents a study of methods based on active learning. The method is based on active learning and combines support vector machine algorithms to construct a learning network similar to the human visual system. Active learning learns the corresponding facial expression action unit through training, and utilizes support vector machines to classify different action units and ultimately maps to corresponding facial expressions, thereby realizing recognition of facial expressions. Experimental results show that this method can effectively suppress correlated noise and obtain relative information. It not only has good robustness, but also improves the recognition rate of facial expressions and can meet actual needs.