Research on Computer Interaction Design Based on User Emotion Recognition Algorithm

Multimodal emotion recognition is a challenging and hot research direction in the field of artificial intelligence. Multimodal emotion recognition mainly improves emotion recognition performance by integrating multiple modal emotion information. The main difficulty lies in whether to learn more discriminative single modal emotional information and whether to fully explore multimodal complementary information through multimodal fusion methods. This article conducts relevant research on expression recognition, multimodal emotion recognition, and human-computer interaction in virtual environments, studies multimodal emotion recognition and its natural interaction in virtual environments, and proposes a model layer fusion method based on neural networks. After fusing speech emotion features and expression features, the complementary information between the fused features is learned through neural networks. Comparative experiments were conducted on the IEMOCAP dataset, and the recognition rate of the model layer fusion method reached 88%. Through comparative verification, it was shown that this method has certain effectiveness.