Emotion recognition method based on pulse neural network multi-modal information fusion
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies suffer from high power consumption in multimodal emotion recognition, and traditional spiking neural networks (SNNs) struggle to achieve adaptive cross-modal spatiotemporal alignment and redundancy suppression, failing to fully utilize the deep complementary relationships between modalities.
A cross-modal dynamic threshold adjustment and collaborative coordinate attention mechanism are introduced. Multimodal information is fused through a spiking neural network. Cross-modal control features are used to back-adjust the firing threshold of cross-attention neurons. A collaborative coordinate attention module is introduced for precise spatiotemporal position encoding.
It achieves low-power emotion recognition, improves the accuracy and robustness of multimodal emotion recognition, can run effectively on resource-constrained edge devices, adapts to individual differences and environmental noise, and improves the accuracy of cross-modal information fusion.
Smart Images

Figure FT_1 
Figure FT_2 
Figure FT_3