ALHCTNet

At present, the micro-expression recognition<br>task has achieved certain results, but maintaining a stable and high<br>recognition accuracy is still a performance bottleneck problem of<br>this task. To address this problem, we propose the Hierarchical<br>CNN-Transformer Network Based on Automatic Localization (ALHCTNet) framework.

This framework leverages three-dimensional<br>optical flow maps to capture information about micro-expression<br>motion sequences, and designs an automatic locator to select local features of five key facial regions in each optical flow map.<br>Considering that the importance of each facial region may vary,<br>we utilize a CNN+Transformer hybrid model. The CNN layer dynamically weights different channels, focusing on those channels<br>beneficial for improving task performance while ignoring unimportant ones.