Emotion Recognition Method Based on Multiscale Attention Residual Network

Facial expression recognition demonstrates broad application prospects in fields such as security monitoring, human-computer interaction, and virtual reality. However, traditional deep learning-based facial expression recognition methods have not fully exploited interchannel correlations and are associated with high computational costs. To address these issues, a facial expression recognition method based on a multiscale attention residual network is proposed. This method introduces a facial multiscale attention module to simultaneously maintain information effectiveness and reduce computational costs through channel grouping. Additionally, a facial feature clustering loss function is proposed to expand the distance between different facial expression feature categories while decreasing the distance between the same facial expression feature categories, making the model training more efficient and robust. Experimental results demonstrate that compared to other mainstream methods, the facial expression recognition method based on the multiscale attention residual network exhibits higher accuracy and robustness, providing valuable insights for research and practical applications in the field of facial expression recognition.