Bimodal Emotion Recognition
Liyanage C De Silva
Abstract
ABSTRACT
In this paper bimodal emo tion recognition has been studied using a statistical technique to classify 6 basic emotions (anger, dislike, fear, happy, sad and surprise) from facial expressions and emotional speech cues recorded in video and audio form. The emotions of 2 human subjects were recorded and analysed by the computer. The emotions anger, sad and surprise had higher success using either video or audio data alone for all the subjects, while the results for dislike were contradicting. The bimodal classification, improves the results in all 6 categories of emotions. When the 6 emotions are grouped into negative and positive emotions classes, it is found that video and audio information are complimentary, and they aid each other in improving the success rates.