Adaptive Convolutional Neural Network-based Information Fusion for Facial Expression Recognition

Ting-Ting Gao ； Hang Li ； Shou-Lin Yin

《International Journal of Electronics and Information Engineering》 13卷1期 (2021/03) Pp. 17-23

https://doi.org/10.6636/IJEIE.202103_13(1).03

摘要

Aiming at the problems of insufficient feature extraction and low recognition rate of convolutional neural network, an adaptive convolutional neural network-based information fusion for Facial expression recognition is proposed. In this paper, the method of gradient feature and texture feature fusion is adopted to effectively combine the local shape feature and local texture feature to perform a more comprehensive representation of facial expression information and obtain different feature information for information fusion at the full connection layer. A convolutional neural network model with two-channel full connection layer is constructed to enhance the characteristic expression ability of the model. The proposed method can recognize 7 basic facial expressions with high accuracy, and has higher recognition accuracy compared with the single-channel facial expression recognition algorithm. Compared with other two-channel convolutional neural networks, it can achieve better recognition effect with simpler network structure.

關鍵字

Double-channel Convolutional Neural Network ； Facial Expression Recognition ； Feature Fusion ； Gradient Feature ； LBP