Facial emotion recognition via stationary wavelet entropy and Biogeography-based optimization

Abstract

INTRODUCTION: As one of the important research directions in the field of computer vision, facial emotion recognition plays an important role in people's daily life. How to make the computer accurately read facial emotion is an important research content. OBJECTIVES: In the current research on facial emotion recognition, there are some problems such as poor generalization ability of network model and low robustness of recognition system. To solve above problems, we propose a novel facial emotion recognition method. METHODS: Our method of feature extraction using the stationary wavelet entropy, which combines single hidden layer feedforward neural network with biogeography-based optimization for facial emotion recognition. RESULTS: The simulation results show that the overall accuracy of our method is 93.79±1.24%. CONCLUSION: This model is superior to the current mainstream facial emotion recognition models in the performance of facial emotion detection. In future research, we will try deep learning and other training methods.

A facial expression recognizer using modified ResNet-152

“…The state-of-art methods are HWT [9], CSO [10], and BBO [42], and the corresponding OA is 78.37+1.50%, 89.49+0.76%, and 93.79+1.24% respectively. The result of comparison is shown in Table 3.…”

Facial expression recognition via transfer learning

“…For the sensitivity and overall accuracy (OA) of the network after the implement of r=10,g=10, we can obtain the following formula to define: 8) is the confusion matrix, is the number of iteratio ns, and is the number of groups.…”

“…In this experiment, the OA of "ResNet-101" combined with transfer learning method was compared with that of PCA + SVM [3], HOS [4], CSO [7] and BBO [8]. The results were shown in Table 3 and Figure 12: the OA of PCA + SVM [3] was 89.14±2.91%, the OA of HOS [4] was 83.43±2.15%, the OA of CSO [7] was 89.49±0.76%, and the OA of BBO [8] was 93.79±1.24%.We can clearly see that the accuracy of "ResNet-101" combined with transfer learning method is the highest (96.29±0.78%), followed by BBO [8], CSO [7], PCA + SVM [3]. It can be seen from Table 3 that the "ResNet-101" method obtains the highest OA mainly depends on: the existence of ResNet-101 residual module solves the problem of network degradation; In the process of deep learning exploration, it is found that the expansion of depth is far better than the expansion of breadth.…”

“…And the structure of Shortcut connection also ensures that when the parameters are updated by backpropagation, it is difficult for the gradient to be 0 and the gradient will not disappear. The second best method is the BBO [8] algorithm, which is developed on the basis of genetic algorithm and particle swarm optimization. It is suitable for solving high-dimensional, multi-objective optimization problems.…”

“…Yang [7] utilized cat swarm optimization to recognize facial emotions. Li [8] chose to use biogeography-based optimization (BBO) for the same task. However, through the analysis of the above methods, it is found that there are some problems such as low accuracy and low robustness.…”