Research on Real-Time Expression Recognition of Complex Environment Based on Attention Mechanism

With the development of deep learning, facial expression recognition has already begun to show results, and facial recognition has a wide range of applications. Whether it is in the field of criminal investigation, education, or medical treatment, it has a bright application prospect. However, in complex situations, due to the influence of factors such as face posture, occlusion, and lighting, facial expression recognition still faces great challenges. In view of the current low accuracy of facial expression recognition in complex situations and the poor real-time performance caused by the diversity and complexity of network structures, this paper proposes a real-time facial expression recognition system based on attention mechanism, which includes separable CNN, residual network, and computer vision attention mechanism. Through the combination of separable CNN and residual network, the number of parameters is greatly reduced, and its real-time requirements are guaranteed. The attention mechanism is used to focus on the detection target and improves the recognition accuracy. Experiments on the face expression dataset of complex scenes in fer-2013 show that the attention mechanism can significantly improve the recognition rate of expressions, and the network also maintains a good real-time effect.