Description
People counting / occupancy estimation predicts the number of humans in a sensing region — as a binary (present / absent), bucketed (0 / 1 / 2-5 / 6+), or fully integer count. It is the most directly thesis-relevant CSI inference task and the one the BLE-calibrated CSI architecture is designed to make robust. Pipelines fall along a spectrum from hand-crafted feature regressors to density-map CNNs.
When it's used
- Smart-building HVAC control
- Real-time occupancy dashboards
- Capacity / safety monitoring
- Calibration target for the BLE-CSI fusion pipeline
Limitations
- Generalisation across rooms / hardware is poor without calibration
- Multi-person separation degrades quickly above 5–6 occupants
- Static people are under-counted by motion-based features