Fall Detection for Elderly: How It Actually Works (and Why Most Fail)

Why fall detection matters more than people realise

Falls are the leading cause of injury-related death in adults over 65 globally, and India is no exception. The dangerous part is not always the fall itself. It is the time-on-ground after. A senior who falls and cannot get up loses dexterity within 30 minutes, develops pressure injuries within 2 hours, and can develop rhabdomyolysis (muscle breakdown that causes kidney failure) within 6 hours. Fast detection is not a nice-to-have; it is what separates a recoverable incident from a fatal one.

The three sensors involved

Modern fall detection uses three sensors in combination. Accelerometer: measures linear acceleration on three axes (x, y, z), the impact spike. Gyroscope: measures rotational velocity, captures the body rotating from vertical to horizontal. Barometer (in advanced devices): measures altitude change: confirms the body actually descended in elevation. Cheap devices use only an accelerometer; this is why they fail.

The physics signature of a real fall

A real fall produces a very specific pattern: free-fall phase (acceleration drops near zero for 200–400 milliseconds), impact spike (sharp acceleration peak above 2g), orientation change (gyroscope shows 90-degree rotation), no-motion phase (person lies still or moves weakly). Algorithms look for this entire sequence, not just the impact. A device that triggers on impact alone will fire false positives every time you slap a tabletop.

False positives: the dirty secret of cheap devices

Cheap fall detectors have false positive rates of 25–40 percent. Things that fire them: vigorous Indian-style hand gestures during conversation, dropping the device on a hard surface, getting on/off a motorcycle, sitting down too quickly, even sneezing while wearing it. The user disables fall detection within a week because of false alerts, and now the safety feature does not exist at all. This is the silent failure mode of most affordable fall detectors.

False negatives: the dangerous failure

Even worse than false positives are false negatives: the device misses a real fall. Soft falls onto a sofa or carpet do not produce sharp impact spikes. Slow slumps from standing (vasovagal episodes, heart attacks) lack the free-fall signature. Falls onto the back versus side versus front have different rotational signatures. A good fall detection system handles all of these; budget devices typically catch only sharp side-falls onto hard surfaces.

Machine learning is what closes the gap

Modern fall detection uses ML models trained on tens of thousands of labelled fall and non-fall sequences. The model takes the sensor data over a 1.5–2 second window and outputs a probability score. Above a threshold, the device triggers the confirmation window. Apple Watch and Galaxy Watch use ML. The Kavach Band uses the maker's sensor-pattern detector, which looks for the whole sequence above rather than the impact alone; we have tested it, not certified it. The cheapest devices use a single impact threshold, which is why their accuracy is poor.

The 30 to 60 second confirmation window

After detection, the device does not immediately call for help. It vibrates, beeps, and shows a "are you OK?" prompt for 30–60 seconds. If the wearer cancels (because they are fine), no alert fires. If they do not cancel, meaning they may be unconscious or unable to move, the device auto-triggers SOS, sends location to emergency contacts, and (depending on device) places an emergency call. This window prevents false alerts while preserving fast response when it matters.

Why the SOS routing matters more than detection

Detection accuracy gets all the attention, but SOS routing decides whether help actually arrives. Questions to ask: who gets notified? (One contact, multiple, emergency services?) How? (SMS, push notification, phone call?) Is location included? (Lat/long, address, live tracking?) Does it work without a phone? (Embedded cellular vs phone-pairing?) Can the wearer cancel a false alert easily? Many devices detect falls well but route the alert poorly.

Indoor accuracy: where most devices struggle

Most falls happen indoors: bathrooms, bedrooms, stairs. GPS does not work indoors. Modern devices use Wi-Fi positioning + cellular tower triangulation + last-known-GPS to estimate indoor location. Accuracy varies. The barometer helps confirm "person fell from standing height" but does not pinpoint the room. For elderly parents living alone, complement fall detection with smart home presence (Alexa motion sensors, smart lights) to identify which room they are in.

Battery life impacts fall detection reliability

Fall detection runs continuously, which drains battery. Smartwatches with fall detection enabled lose 15–25% extra battery. This is why many users disable fall detection on Apple Watch, to extend battery from 14 hours to 18. A device that needs daily charging often gets removed at night, exactly when bathroom falls peak. This is a real tension for every device in this category, Kavach included: a charge lasts up to 24 hours and around ten hours of that with GPS reporting every ten minutes, so the honest answer is to charge it while the wearer is asleep and accept that the hours on the charger are uncovered.

How Kavach approaches fall detection

Kavach uses motion sensing on the band to flag a fall, then gives the wearer a 30-second window to cancel before an SOS goes to up to three contacts over the 4G connection built into the band. We are still validating detection accuracy in the field and have not published a figure, because we do not yet have one we would stand behind. The step-by-step, and the list of what it can get wrong, is on the fall detection page.

How to choose a fall detection device

Five checks: (1) has the maker published a detection and false-positive rate, and from what sample size? (2) what is the SOS routing, and does the alert carry a live location? (3) does it work with the wearer's phone switched off? (4) what is the battery life with fall detection enabled, measured in use rather than standby? (5) what happens when the device is taken off the wrist? Ask for figures, and treat "we do not have that number yet" as a more trustworthy answer than a round one with no source. On question one we are in the first category ourselves, and we would rather say so than invent a percentage.

For Indian families specifically

India has unique considerations: parents living alone in tier-2/3 cities while children are in metros, two-wheeler riding common, multi-generational homes with stairs, bathroom flooring that is wet and slippery. Kavach is built for these realities. Reserve at www.ourkavach.com/pre-order: ₹5,999 with 3 months of the Safety Plan included. The optional Safety Plan is ₹149/month for live tracking, route history and geo-fences; SOS and fall alerts work free for the life of the band.

Reserve your Kavach Band

Real-time GPS, one-touch SOS and fall detection, on its own 4G connection. Starting ₹5,999, no payment required to reserve.

Related reads