Multimodal Affective Computing in Employee Care Based on the Multi-Fusion Residual Memory Model

The application of artificial intelligence in the field of affective computing has been expanding, especially in enterprise employee care, and the multimodal data fusion model has become the key to improve the accuracy of emotion recognition. The study proposes a multimodal residual memory network model, which adopts multimodal data inputs such as visual and physiological signals, combines gated cyclic units for temporal modelling, and uses the affective intensity attention mechanism for feature fusion, as well as a bi-directional long and short-term memory network for time-step level affective modelling, and finally adopts a cross-view learning module for classification. The experimental results show that the model achieves an emotion recognition accuracy of 91% with 500 data samples, and the root mean square error is reduced to 0.1, which is better than the other compared models. For the five emotion categories, the model has an average recognition accuracy of 71%, which is an improvement of about 12% compared to the model based on the multifactor hierarchical framework. The results show that the method has strong potential for application in the field of multimodal affective computing, and can provide more accurate emotion recognition support for employee care systems.