Facial Expression Recognition of Video Conference Participants Based on Improved 3D Convolutional Neural Network Algorithm

With the increasing popularity of remote communication forms such as video conferencing, facial expression recognition of participants has become an important research direction in intelligent video analysis. However, traditional two-dimensional convolutional neural networks have certain limitations in processing time series information, and their ability to capture dynamic temporal features in video data is insufficient. A three-dimensional convolutional neural network algorithm has been introduced to address this issue. Meanwhile, to enhance the attention of important information in the feature map, a convolutional block attention module was used to optimize the original network, and the improved algorithm was applied to facial expression recognition of video conference personnel. The results showed that during the training process of each model, the algorithm combining convolutional block attention module and 3D convolutional neural network achieved the highest accuracy of 96.54%. At the same time, the recognition accuracy, recall rate, and $\mathbf{F 1}$ score of the algorithm during the testing phase were as high as $95.82 \%, 93.61 \%$, and 93.75%, respectively. In addition, the area under the receiver operation characteristic curve of the algorithm proposed by the research institute is as high as $\mathbf{0. 8 5 6 4}$. The improved algorithm proposed in the study has significant recognition performance and can effectively recognize facial expressions in video conferences in complex backgrounds and long time series, providing effective technical support for future intelligent video analysis. In the future, this technology has potential advantages in multiple fields such as sentiment analysis, mental health assessment, and user experience optimization.