How to detect falls in a care home

Rounds are hourly at best. A fall takes a second. The gap between those two numbers is the thing everyone who runs a care home already worries about: a resident on the floor between checks, waiting to be found. Nobody needs statistics to feel the weight of that, and this page won't invent any. The question is simply how to shorten the wait.

Why wearables only half work

Pendants and wrist alarms ask three things of a resident: wear the device, keep it charged, and press it at the worst moment of their week. The residents most at risk of falling are often the least able or willing to do all three, and a pendant left on the bedside table protects nobody. A camera watches the room rather than the device, so there's nothing to forget, remove or reach for.

How camera-based detection works here

Svid has a vision model read frames from the cameras a home already has, describing what's happening in plain English. Fall / Slip Detection, one of the alerting checks, judges a window of recent frames rather than a single snapshot, so a stumble that ends with someone on the floor reads as an event, and the alert goes to staff at once rather than into a report later. Bed Exit Alert covers the moment before: a resident who shouldn't be up unaided has left the bed, and staff know before the fall rather than after it. Rooms where falls are the worry suit the Precise watch level, a frame every second, because a fall doesn't wait five seconds to happen.

1. 01

   Connect the cameras you already have

   Corridors and communal rooms are the usual starting point, over RTSP with no new hardware. Whether cameras belong in bedrooms is a privacy judgement for the home to make, with residents and families part of that conversation.

2. 02

   Switch on the right checks, where they're needed

   Fall / Slip Detection where falls happen, Bed Exit Alert in the rooms that need it. Each camera gets exactly the watching it needs and nothing more, and each check is priced in the Store from the home's own usage.

3. 03

   Route alerts to the people on shift

   On screen, by email, or by webhook into an existing messaging or nurse-call system, so the alert lands where staff are already looking.

4. 04

   Review incidents properly afterwards

   The footage is searchable in plain English, so “what happened before the fall” is a question you can actually answer, and handover notes can draw on a written Daily Summary report rather than memory.

Dignity and privacy

There is no facial recognition, no biometric ID and no tracking of residents, ever. The only AI in the product describes what's visible in a frame and checks conditions on it. In practice that means events are flagged, residents aren't identified: the alert says someone has fallen in the day room, not who. And for footage this sensitive, a fully on-prem appliance mode exists, running on the home's own infrastructure, so footage of residents never has to leave the building. In a setting where trust is the whole job, these aren't small details.

Do residents need to wear anything?

No. The cameras watch the room, so there's nothing to wear, charge or press, and nothing that stops working because it was left on a bedside table.

How quickly do staff find out about a fall?

As it happens. Fall / Slip Detection judges frames as they arrive and raises the alert the moment the sequence reads as a fall, on screen, by email or by webhook, not in a summary at the end of the day.

Is it recording residents with facial recognition?

No. There's no facial recognition, no biometric identification and no tracking of residents anywhere in Svid. The AI describes what's happening in the frame and raises the alert; events are flagged, people aren't identified.

Can it tell us someone is out of bed before they fall?

Yes. Bed Exit Alert tells staff when a resident who shouldn't be up unaided leaves the bed, which often means help arrives before a fall instead of after one.