Data augmentation in WiFi/CSI sensing research refers to the process of artificially expanding training datasets by generating new synthetic samples or applying transformations to existing CSI measurements, thereby increasing data diversity without the cost of additional physical data collection. This is particularly critical in the field because CSI signals are highly sensitive to environmental conditions, antenna placement, and domain shifts, meaning models trained on limited real-world data frequently fail to generalize across different rooms, device configurations, or users. Key variants include physics-based generative approaches, such as NeRF²'s use of neural radiance field simulation to synthesize RF propagation data across novel spatial configurations, and topology-aware transformation strategies, such as those employed in WiGNN, which exploit the dynamic spatial relationships among receivers to produce augmented graph-structured inputs that better capture cross-domain variability in gesture recognition tasks.
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
- NeRF2: Neural Radio-Frequency Radiance Fields ↗ — NeRF2: Neural Radio-Frequency Radiance Fields
- RSSI-Assisted CSI-Based Passenger Counting with Multiple Wi-Fi Receivers ↗ — RSSI-Assisted CSI-Based Passenger Counting with Multiple Wi-
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired