Expression recognition method and system based on multi-scale features and spatial attention
By combining the HNFER neural network model with the CoordAtt and SAFM mechanisms, the difficulties of facial expression recognition in complex scenarios using traditional methods are solved, achieving highly accurate and robust facial expression recognition that is suitable for applications such as human-computer interaction and health diagnosis.
Patent Information
- Application Number
- PCT/CN2024/135203
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-06-04
- Filing Date
- 2024-11-28
- Publication Date
- 2025-12-11
AI Technical Summary
Existing technologies are ineffective in handling complex scenarios such as varying lighting, facial expression intensity, and facial occlusion. Traditional facial expression recognition methods struggle to accurately capture and process subtle changes in expression and lack dynamic attention to local features.
An expression recognition method based on multi-scale features and spatial attention is adopted. Feature extraction is performed through the HNFER neural network model, and the CoordAtt and SAFM mechanisms are combined to enhance the feature extraction and classification capabilities, including multi-scale feature extraction, spatial attention mechanism and deep convolution processing.
It improves the accuracy and robustness of facial expression recognition, maintains high performance in complex environments, and is suitable for fields such as human-computer interaction, security systems, and health diagnosis.
Smart Images

Figure CN2024135203_11122025_PF_FP_ABST
Abstract
Citation Information
Patent Citations
Multi-scale feature extraction and attention mechanism fused expression recognition method
CN115909455A
Facial expression recognition method based on space attention network
CN116092146A
Facial expression recognition method and system based on feature fusion and attention mechanism
CN116189272A
Multi-person real-time facial expression recognition method and system based on attention mechanism
CN117636436A
Expression recognition method and system based on multi-scale features and space attention
CN118298491A