Audiovisual emotion recognition

It is a fact that human emotions are expressed in the voice as well as on the face. Early studies showed that combining these two modalities could improve the recognition accuracy. In this paper we have extensively researched the available audio and video based emotion recognition techniques as well as bimodal techniques and then combined the audio features and video features using a rule-based technique to get improved recognition results. Experiments showed that negative emotions such as anger and sadness could be recognized from each other with almost 100% accuracy by using audio features. Audio-based recognition dropped to 93.6% accuracy if the recognition of joy and surprise is included along with the angry emotion. We have observed that the proposed rule-based combined audiovisual emotion detection technique improves the recognition accuracy of all six universal emotions even if considered them together