A multi-modal emotion recognition method based on facial multi-scale features and cross-modal attention (MS-FCA) network is proposed. The MSFCA model improves the traditional single-branch ViT network ...