Multi-Channel Expression Recognition Network Based on Channel Weighting

Accurate expression interpretation occupies a huge proportion of human-to-human communication. The control expressions can facilitate more convenient communication between people. Expression recognition technology has also been transformed from relatively mature laboratory-controlled research to natural scenes research. In this paper, we design multi-channel attention network based on channel weighting for analysis in scenes. mainly consists three parts: Multi-branch feature extraction network, which combines residual ResNet18 and ConvNeXt ideas improve uses adaptive fusion build complete network; Adaptive Channel Weighting, designs weights the auxiliary extraction, performs weighting, highlights key information areas; Attention module, modifies spatial mechanism increases accelerate acquisition important areas. experimental results show that proposed method achieves better efficiency than existing algorithms dataset FER2013 under uncontrolled conditions, reaching 73.81%, good accuracy 89.65% 85.24% Oulu_CASIA RAF-DB datasets, respectively.