Domain generalization in WiFi/CSI sensing refers to the ability of a trained model to maintain reliable performance when deployed in conditions that differ from those seen during training, such as changes in environment, antenna placement, user identity, or the passage of time, without access to target-domain data for retraining or fine-tuning. It matters for the field because CSI signals are highly sensitive to physical context, meaning models that overfit to training conditions fail to transfer to real-world deployments where such conditions inevitably vary. Key variants include temporal generalization, which addresses performance degradation as channel conditions drift over time, and cross-domain generalization, which targets variability across different spatial configurations, hardware setups, or user populations, as exemplified by cross-environment and cross-user gesture recognition scenarios.
Source Papers
- CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing ↗ — CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- Time matters: Empirical insights into the limits and challenges of temporal generalization in CSI-based Wi-Fi sensing ↗ — Time matters: Empirical insights into the limits and challen
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired