Description
WiFi CSI sensing is the concrete realisation of wifi-sensing that uses Channel State Information (per-subcarrier complex channel response) as its observation, rather than scalar RSSI. It is the core method family of this thesis — every Chapter 4 (CSI sensing) experiment lives inside this category. Pipelines typically convert raw CSI into amplitude / phase / ratio features and feed them to learned models for occupancy, motion, or activity inference.
When it's used
- All Chapter-4 CSI experiments in the thesis testbed
- Crowd-counting and density-estimation studies
- Activity / gesture / gait recognition on commodity APs
- Baselines compared against future BLE-calibrated CSI methods
Limitations
- Inherits commodity-NIC constraints (Intel 5300, Atheros, Nexmon, ESP32)
- Environment dependence is acute — models rarely transfer between rooms
- Standardisation only emerging via 802.11bf