Research and Analysis of Facial Expression Recognition Based on Deep Neural Networks

Yali Yu

doi:10.54097/bhqat185

Authors

Yali Yu

DOI:

https://doi.org/10.54097/bhqat185

Keywords:

Expression Recognition, Neural Network, Depthwise Separable Convolution

Abstract

Since the traditional feature extraction algorithm cannot extract a large number of effective high-dimensional expression features, and the traditional convolutional network model has a large number of parameters in expression recognition and weak generalization ability, this paper selects the deep convolutional neural network Xception architecture as the basis for improvement. The core operations in the model are residual module and depthwise separable convolution, and ReLU6 activation function is used. The improved model is trained and tested using the public dataset CK+. Through multiple training experiments, the results show that the improved model has achieved a certain level of facial expression recognition performance.

Downloads

Download data is not yet available.

References

[1] Ekman P, Friesen W V. Constants across cultures in the face and emotion[J]. Journal of Personality and Social Psychology, 1971, 17(2): 124.

[2] Dang Xin, Xu Hua. Driving state Analysis based on Facial Expression Recognition [J]. Journal of Information Recording Materials, 2019,25(3):108-111,114.

[3] Li Jing, Li Jian, Chen Haifeng, et al. Facial expression recognition based on occlusion and reconstruction of key areas [J]. Computer Engineering, 2019,50(5):241-249.

[4] Li Chunhong, Lu Yu. Facial expression recognition based on depth-separable Convolution [J]. Computer Engineering and Design, 2019,42(5):1448-1454.

[5] Liu Jin, Luo Xiaoshu, Xu Zhaoxing. Lightweight Facial expression recognition using spatial grouping to enhance attention [J]. Computer Engineering and Applications, 2019, 59 (22):233-241.

[6] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 1251-1258.

[7] Chen L, Peng L, Yao G, et al. A modified inception-ResNet network with discriminant weighting loss for handwritten chinese character recognition[C]. 2019 International Conference on Document Analysis and Recognition (ICDAR). IEEE, 2019: 1220-1225.

[8] Opschoor J A A, Petersen P C, Schwab C. Deep ReLU networks and high-order finite element methods[J]. Analysis and Applications, 2020, 18(05): 715-770.

[9] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 4510-4520.