Facial Recognition: Decoding Emotions in Online Collaboration

Following the rapid increase in online communication and remote collaboration after the COVID-19 pandemic, there has been a heightened need to understand and manage emotional dynamics in virtual interactions. This study constructed and validated an online real-time emotion recognition system. Utilizing FaceMesh for facial landmark detection, the Fer2013 dataset for training, and the Xception deep learning architecture for emotion classification, the system efficiently deciphers emotions from real-time videos. It was observed that participants in two different groups showed a preference for a free discussion format when engaged in two distinct discussion modes. This observation aligns with the results from participant surveys, confirming the accuracy of the system of emotion predictions. The construction and validation of this system lay the foundation for the subsequent development of active communication model.