A Study on Emotion Recognition Based on Hierarchical Adaboost Multi-class Algorithm

Researches on human emotion recognition have attracted more and more people's interest. Adaboost algorithm is an integrated algorithm that constructs strong classifiers by iterative aggregation of weak classifiers. This paper proposes a hierarchical Adaboost (HAdaboost) multi-class algorithm for emotion recognition, which improves the original Adaboost algorithm. The valence and arousal in different emotional states are used as classification features, and emotion recognition is performed according to their differences. Simulation experiments on the Chinese Facial Affective Picture System (CFAPS) data set demonstrate three types of emotions and seven types of emotions can be distinguished, and the average accuracy rates are 93% and 92.4% respectively.