If you are specifying sensors for elderly-care monitoring and someone proposes fusing a Time-of-Flight (TOF) depth sensor with a mmWave radar module, here is the direct answer: the combined precision is genuinely higher — not as a slogan, but because the two technologies fail in different places and cover each other's blind spots. This article states the conclusion first, then shows the mechanism, where fusion actually pays off, where it does not, and how cost should drive the final choice.
TOF: optical ranging with a hard ceiling
TOF is fundamentally an optical method — infrared, laser, or similar. It emits light, measures the round-trip time, and computes the distance to whatever reflected it. Put enough measurement points together and you reconstruct a depth map: the contour of a body, head, shoulders, limbs. The distance resolution is excellent; you can tell how far each part is with near-sub-centimetre clarity.
The ceiling is physical, not a tuning issue: light does not penetrate. Fog, water vapour, or a layer of clothing between the sensor and the body scatters or blocks the return, and the sensor simply cannot tell what is behind the obstruction. Every optical approach shares this limitation.
mmWave radar: motion and penetration, weak at counting
A mmWave radar is exquisitely sensitive to motion. It detects whether something is moving, how much, and at what micro scale — breathing-level movement included. Its radio energy penetrates clothing and, to a useful degree, water vapour and steam, so it can judge whether a person is truly lying on the floor and whether their motion has changed, even where an optical sensor would be blinded.
The weak spot is people separation. Two bodies standing close together merge into one blurred cluster of motion points; the radar cannot resolve a clean contour of each person the way a depth camera can.
Why fusion is more than the sum
| Capability | TOF (optical) | mmWave radar | Better together |
|---|---|---|---|
| Distance / contour resolution | Excellent (sub-cm depth map) | Poor (no clean contour) | TOF leads |
| Motion & micro-motion (breathing) | Weak (needs visible motion) | Excellent | Radar leads |
| Penetration (clothing, vapour) | Poor (blocked) | Good | Radar leads |
| People counting / separation | Good (distinct contours) | Poor (blurs when close) | TOF leads |
| Robustness in steam / fog | Poor | Good | Radar leads |
One sensor is good at distance and shape; the other at motion and penetration. Stack them and each covers the other's failure.
The bathroom fall-alarm scenario
This is where the complementarity is most obvious — and it is exactly the place fall alarms matter most. A bathroom is where optical sensors struggle: steam, splashing water, a body partly behind a shower screen or under a towel.
- TOF is defeated by water vapour and occlusion — it cannot see through the steam or past the screen.
- mmWave penetrates clothing and vapour, and reports whether the person has really gone down and whether motion has stopped or changed — the questions a fall alarm actually asks.
- But the same bathroom can hold two people (a carer and the resident). For counting and telling them apart, TOF recovers clean separate contours where the radar would blur them into one.
Net result: fusion is more accurate for fall detection (radar sees through what blinds TOF) and for people counting (TOF resolves what radar blurs). Both failure modes that plague single-sensor installs get closed at once.
Cost and product positioning
TOF is not free to add. It needs an optical lens, careful optical stacking, and tighter mechanical design; that complexity raises both the bill of materials and the assembly cost. A mmWave radar is comparatively simple — often a single module solves the sensing problem.
So fusion is not automatically the right answer for every product. The precision gain is real, but it costs more and is heavier to design. Selection should follow the use case and the population: a high-stakes bathroom fall alarm for a frail person justifies the fused precision, while a simple presence-or-not lamp in a bedroom may be perfectly served by a plain radar module. Match the sensor bill to the risk.
Where fusion does NOT help
Fusion raises precision only where the blind spots are complementary. It does nothing for:
- Wrong geometry or poor install — if neither sensor is aimed at the scene it watches, fusion just combines two wrong answers.
- Calibration and sync — the two sensors must be aligned and time-synced; drift or disagreement between them creates a new error class (which sensor do you believe?).
- Compute and power — running and fusing both streams costs more processing, and often more power, than either alone.
And the standing disclaimer from the rest of this series: this is not a medical device. Fusion improves the sensor's judgement; it does not replace human checks. It publishes state to your own system, and what happens next is what you build.
Takeaway
- TOF gives contour and distance; radar gives micro-motion and penetration. They fail in different places.
- Fusion is genuinely more accurate where those failures overlap — especially bathroom fall detection and people counting.
- TOF is defeated by steam and occlusion; radar penetrates them. Radar blurs close bodies; TOF separates them.
- Fusion costs more (optics, stacking, compute) — justify it by risk, not by default.
Fusion is not magic: wrong install, calibration drift, and power budget still apply. It is an addition to a care routine, never a substitute.
See also: why naive radar failed four bathroom scenes and how it was solved, our presence detection radar: range, sensitivity, scene modes, and why our fall radar is tuned for bathrooms.
If you are building with these sensors, the hardware we use is on Tindie:
Check out our mmWave radar modules for elderly-care and fall detection
Top comments (0)