An attention mechanism is a neural network component that dynamically assigns importance weights to different parts of an input sequence or feature space, allowing the model to focus selectively on the most task-relevant information rather than treating all inputs equally. In WiFi CSI sensing, attention is critical because raw CSI signals are inherently noisy and high-dimensional, containing subcarrier, temporal, and spatial components of varying relevance to activities or gestures, and attention helps the model suppress interference while amplifying discriminative patterns. Key variants employed in this domain include self-attention and multi-head attention (as used in Transformer-based architectures benchmarked in SenseFi), as well as graph attention mechanisms (as in WiGNN), where attention weights are computed over dynamic graph topologies to capture the relative importance of spatial links between distributed WiFi receivers during cross-domain gesture recognition.
Source Papers
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- RSSI-Assisted CSI-Based Passenger Counting with Multiple Wi-Fi Receivers ↗ — RSSI-Assisted CSI-Based Passenger Counting with Multiple Wi-
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired