Research on Emotion Recognition for Online Learning in a Novel Computing Model

The recognition of human emotions is expected to completely change the mode human-computer interaction. In emotion research, we need focus on accuracy and real-time performance in order apply emotional based physiological signals solve practical problems. Considering timeliness dimension recognition, propose a terminal-edge-cloud system architecture. Compared traditional sentiment computing architectures, proposed architecture this paper reduces average time consumption by 15% when running same affective process. Proposed Joint Mutual Information (JMI) feature extraction model, conducted extensive experiments AMIGOS dataset. Through experimental comparison, network has obvious advantages over commonly used methods. model performs classification, valence arousal 71% 81.8%, compared with recent similar classifier improved 0.85%. addition, set up an experiment 30 people online learning scenario validate algorithm model. result proved that were satisfactory, interaction experience.