← WiFi Sensing Fundamentals

Beyond presence: multi-modal sensing

10 min

When one sense is not enough

CSI answers "is someone there, are they moving?" Real problems often need more: Is that motion a person or a pump? Is the machine healthy? How far away? The answer is fusing cheap modalities — the Multi-Modal Sensor Hub in the catalog carries five.

The fusion vector

Our field configuration produces a 510-dimension feature vector per measurement:

  • WiFi CSI — 56 subcarriers: presence, motion, through-wall reach.
  • Microphone FFT — 256 bins: breathing band (0.2–0.5 Hz envelope), machine harmonics, glass-break transients.
  • IMU FFT — 32 bins: structural vibration — bearing wear and cavitation have signatures long before failure.
  • Ultrasonic — 128 bins: line-of-sight distance + motion, immune to radio interference.

Each modality covers another's blind spot: ultrasonic needs line of sight, CSI does not; CSI cannot hear a bearing, the IMU can.

Real deployments of this exact stack

  • Data-center cages — CSI mesh detects wrong-cage and after-hours entry where cameras are contractually banned.
  • Oil & gas hazardous zones — the full 510-D stack watches for people in H₂S areas and listens to rotating equipment, where electronics must be minimal and cameras are prohibited.
  • RF-triggered evidence — CSI motion wakes a camera for a 10-second burst (CameraPlus): privacy-preserving and energy-saving.

Design rule

Start with the question, not the sensors. "Alert when someone enters after hours" is CSI alone. "Alert when the pump sounds wrong" is mic + IMU. Buy sensors for the question you actually asked — the hub is there when the question grows.


Sign in to use this feature — it takes 20 seconds and it’s free. Sign in Back to course