Emotion Classification and Recognition based on facial EMG

Abstract

∗ Study emotion classification recognition individual difference is big, dispersion characteristics of rule, the problem of insufficient accuracy, based on the physiological signal acquisition device collected 12 subjects zygomatic muscle and brow muscle two channels of electromyography data, from the time domain feature extraction, using support vector machine (SVM) and the method of extreme learning machine(ELM) to classify positive, negative and neutral moods,Compare the classification accuracy and find out the algorithm with higher classification accuracy. The results showed that the zygomatic muscle activity increased significantly and the frowning muscle activity decreased significantly in positive emotions, while the frowning muscle activity increased significantly and the zygomatic muscle activity decreased in negative emotions.Compared with the 50% average classification accuracy of support vector machine classifier, the average classification ef-ficiency of extreme learning machine classifier is better, and the average classification accuracy can reach 60.08%.In practical ap-plications, the extreme learning machine has a better classification effect and provides a certain technical foundation for modern human-computer interaction.