Research on Application of Un-supervision Deep Learning in Nonspecific Facial Expression Recognition

Abstract

The traditional deep learning model has some defects, such as low recognition accuracy and long recognition time. Therefore, unsupervised algorithm is introduced to recognize non-specific facial expressions. Firstly, the non-specific facial expression image was preprocessed, including clipping, scaling and gray normalization. Secondly, according to the preprocessing results, the unsupervised deep learning model was constructed by combining the deep convolution neural network and unsupervised learning. Finally, the recognition of nonspecific facial expression was achieved. The results show that the unsupervised deep learning model has high average recognition accuracy and short recognition time