Facial Emotion Recognition Based on CNN

With the development of artificial intelligence, computers will have not only IQ but also EQ in the future. Affective computing, which makes computers have emotion, has received more and more attention in recent years. Among them, facial expression recognition has become a research hotspot in the field of affective computing. In this paper, facial expression recognition is studied based on Valence-Arousal dimensional emotion model. A facial expression valence dimension prediction system based on convo-lution neural network is designed in this study. The system in-cludes face detection, feature extraction, valence grade prediction and so on. In this system, the annotation of facial expressions is divided into 9 levels. and the probability of each valence dimen-sion is obtained through the output of CNN network, and the final prediction result is equal to the weighted fusion of valence value and its corresponding probability. We use CK+ database and Fer2013 database to complete the training of CNN network model, and verify the performance of the system by recognizing the facial expressions of volunteers when watching video. The results show that the system can correctly predict the emotional effect value of volunteers, and the average RMSE index is 0.0857±0.0064.