Book a Demo
Detect a Fall in Seconds — Not at the Next Round

AI Fall Detection System

The danger of a fall is rarely the impact — it's the time spent on the floor afterwards. A patient falls two minutes after a nurse leaves. A resident falls alone at 3 a.m. A worker slips on an oily floor with nobody nearby. Every minute before someone notices increases the harm. Tentovision's AI fall detection system detects falls and post-fall immobility through the cameras you already have — alerting staff immediately, with privacy modes designed for sensitive environments. (Performance, alert routing & privacy configuration are confirmed for your site at assessment.)

SecondsTo alert
No WearableNothing to press
Privacy ModesEdge · anonymised
Hero image — calm, anonymised treatment
(Fall detected · Ward 3 corridor · staff alerted)
AI fall detection alerting nursing staff to a fall and post-fall immobility in a hospital corridor, using an existing camera with privacy-preserving anonymised output
Nothing to Wear or Press
No Facial Recognition
Runs on Existing Cameras
Post-Fall Immobility
Not a Medical Device
The Problem

The Real Danger Isn't the Fall. It's the Time on the Floor.

Most falls are survivable. What turns a fall into a serious event is how long someone lies there before help arrives — the clinical term is a long lie, and the harm compounds quietly: dehydration, pressure injury, hypothermia, and a delay in treating whatever caused the fall.

In Hospitals, the Gaps Are Structural

Patient falls are a reportable quality indicator, and every ward already runs a falls-prevention programme. But night shifts are the thinnest staffed, and hourly rounding — however diligent — leaves fifty-nine minutes between checks. A patient can fall two minutes after a nurse leaves. That isn't a failure of care; it's arithmetic.

In Care Homes, Residents Are Alone by Design

Independence and dignity mean private rooms, and private rooms mean nobody is present when a resident gets up at night and falls. The resident who most needs help quickly is frequently the one least able to summon it.

In Industry and Kitchens, the Risk Is Distance

Wet or oily floors, cluttered aisles, and lone workers in remote parts of a plant, warehouse, or kitchen mean a person can go down in an area no colleague passes for a long time.

Wearables Help — Until the Moment They're Needed Most

A pendant or pull-cord only works if the person is wearing it, is conscious, can reach it, and can press it. Devices get left on bedside tables, batteries run flat, and some people decline them because wearing an alarm feels like a loss of independence.

Definition

What Is AI Fall Detection?

AI fall detection is a computer-vision system that detects when a person falls — and whether they remain down afterwards — from camera feeds, using body-posture analysis, and alerts staff or caregivers immediately, without requiring the person to wear or activate any device.

The fall event and the immobility that follows are two different things, and the second matters more. Detecting the moment of a fall triggers the alert. Detecting that the person has stayed down distinguishes a genuine emergency from someone who slipped, got up, and carried on — and it's what prevents the long lie.

Passive, not active

Active detection depends on the person doing something: pressing a button, pulling a cord, wearing a device that's charged and correctly worn. Passive detection asks nothing of them. For someone unconscious, disoriented, in pain, or unable to reach a cord, that difference is the whole point.

Posture, not appearance

The system analyses a person's position relative to the floor and the pattern of the movement — geometry rather than identity. That's what makes the anonymised privacy modes below technically possible, rather than a marketing claim bolted onto a conventional camera system.

Where it fits

Fall detection runs on the Tentovision AI Video Analytics platform and its VMS, hosted on AWS with on-site edge options. In commercial kitchens it's one of the modules in our smart kitchen monitoring bundle.

What it is not

It is not a medical device, and it does not diagnose, treat, or predict anything clinical. It does not replace your falls-prevention programme, your rounding, or clinical judgement. It is not staff-productivity monitoring, and it is not surveillance of the people it protects — a distinction the privacy architecture below is built to make real rather than rhetorical.

Capabilities

Capabilities

Capability 01

Real-Time Fall Detection

The point of passive detection is that it asks nothing of the person who has fallen.

Detects a fall from body posture and motion — no wearable, button, or cord required
Works for people who are unconscious, disoriented, or unable to call for help
Calibrated per room or zone, because every space produces different movement patterns
Fall detected from body posture
on an existing camera
Real-time AI fall detection from an existing camera using body-posture analysis with anonymised output
Capability 02

Post-Fall Immobility Detection — the "Long Lie" Safeguard

This is the capability that matters most clinically, because time on the floor — not the fall itself — is usually what causes lasting harm.

Confirms whether a person has remained down after a fall
Distinguishes a genuine emergency from a stumble someone recovers from unaided
Escalation threshold set with your clinical team, not a vendor default
Post-fall immobility —
the long-lie safeguard
Post-fall immobility detection confirming a person has remained down after a fall
Capability 03

Privacy-Preserving Modes

Configured with you, documented for your ethics committee, and confirmed before go-live.

Edge processing, so video need not leave the premises
Silhouette or skeleton-only output where identifiable video is unacceptable
Event-only retention, no audio, no continuous live viewing, and zone exclusions
No facial recognition — posture and movement, never identity
Anonymised skeleton output
instead of identifiable video
Privacy-preserving fall detection using anonymised skeleton output instead of identifiable video
Capability 04

Instant Multi-Channel Alerts

An alert only helps if it reaches the right person where they're already looking.

Alerts within seconds of a fall being confirmed, with the location included
Routed to the channels your team already uses — nursing station, app, SMS, or WhatsApp
Can be configured to integrate with an existing nurse-call or alarm system (scoped per site); escalation can run through monitored e-surveillance
Alert at the nursing station
with location details
Fall alert delivered to a nursing station with location details
Capability 05

Incident Logging & Fall Analytics

It supports your patient-safety reporting and quality review; it doesn't replace your falls-prevention programme or certify any standard.

A timestamped record of each detected fall: when, where, and how quickly it was responded to
Patterns over time — which areas, wards, or hours carry the most risk
Access is role-based and every review is logged
Incident log & fall analytics —
areas, wards, hours
Fall incident log and analytics showing higher-risk areas and times
Works With Your Existing Cameras

Vendor-Agnostic · 200+ Camera Brands Supported

Tentovision connects to any IP or CCTV camera over ONVIF, RTSP, and standard NVR APIs, streaming via the Cloud Adapter where needed — in corridors, wards, common areas, and industrial floors, suitable cameras often already exist. (Camera suitability for fall detection in a given room — angle, mounting height, and lighting — is confirmed at assessment; general-purpose CCTV is not automatically suitable for every space.)

HikvisionWorld #1Hikvision camera brand
DahuaGlobal Top 2Dahua camera brand
CP PlusIndian LeaderCP Plus camera brand
UniviewTier-1 ChinaUniview camera brand
AxisSwedishAxis camera brand
BoschGermanBosch camera brand
HikvisionWorld #1Hikvision camera brand
DahuaGlobal Top 2Dahua camera brand
CP PlusIndian LeaderCP Plus camera brand
UniviewTier-1 ChinaUniview camera brand
AxisSwedishAxis camera brand
BoschGermanBosch camera brand
HanwhaKoreanHanwha camera brand
HoneywellAmericanHoneywell camera brand
VivotekTaiwaneseVivotek camera brand
PelcoAmericanPelco camera brand
MatrixIndian OEMMatrix camera brand
TiandyChineseTiandy camera brand
HanwhaKoreanHanwha camera brand
HoneywellAmericanHoneywell camera brand
VivotekTaiwaneseVivotek camera brand
PelcoAmericanPelco camera brand
MatrixIndian OEMMatrix camera brand
TiandyChineseTiandy camera brand
How It Works

How AI Fall Detection Works

Four steps, running on the cameras you already have. (Calibration, alert routing & integrations are configured per room or zone and confirmed at assessment.)

STEP 01

Posture Analysis

The AI analyses body posture and motion — a person's position relative to the floor — rather than appearance, which is what allows the privacy modes above.

STEP 02

Distinguishing a Fall

It differentiates a fall from sitting down quickly or lying down deliberately, calibrated per room — a lounge chair, a hospital bed, and a warehouse aisle produce very different patterns.

STEP 03

Confirming Immobility

The system confirms whether the person has remained down — the check that separates an urgent alert from a momentary stumble.

STEP 04

Alert & Log

Staff are alerted immediately through the channels they already use, with the location, and the event can be logged for review — in the command centre with AWS cloud or on-site options, and camera health monitoring keeping cameras online.

Privacy by Design

Built for Places Where Privacy Isn't Negotiable

If your instinct is that identifiable video of patients or residents shouldn't sit on a server, that's the right instinct — and it's the design constraint this system is built around.

In healthcare and care settings, privacy isn't a feature to be traded against safety; it's the condition for being allowed through the door at all. Here is how the system can be configured — each mode is confirmed per deployment, and where one isn't available for your setup, we'll tell you plainly rather than imply it.

Edge processing

Analysis can run on-site, so video need not leave your premises for a fall to be detected. What travels is the alert, not the footage.

Silhouette & skeleton-only output

Where identifiable video is unacceptable, the system can work from an anonymised representation of body posture — geometry, not appearance.

Event-only retention

Rather than recording continuously, it can retain nothing during normal activity and keep only a short clip around a detected fall.

No audio, no continuous live viewing

Fall detection does not require listening, and it does not require anyone to be able to watch a live feed of a private room.

Zone exclusions

Bathrooms, changing areas, and any space your clinical leadership designates are configured out — not blurred, not de-prioritised, but excluded.

No facial recognition

The system detects posture and movement, not identity. It does not attempt to recognise who a person is.

Role-based access, audit logs, and a defined retention period

Only authorised staff can review a fall clip, every access is logged, and clips are held for an agreed period and then removed.

Approaches Compared

Camera-AI vs Wearables vs Radar vs Floor Sensors

No single approach is right for every space, and being straight about that matters more here than in most categories. (Specifics are typical — verify for your setting.)

Dimension Camera-AI (Tentovision)Existing cameras Wearable / pendantWorn device Radar sensorCamera-free Floor / bed sensorSingle point
Nothing to wear or press✗ requires wearing + pressing
Works if the person is unconscious~
Post-fall immobility
Privacy profile~ needs configured modes✓ camera-free
Bathroom suitability✗ not recommended✓ their strength~
Works outside the building✓ travels with the person
Coverage area✓ corridors, wards, floors, plant areas~ person only~ per room✗ single point
Uses existing infrastructure✓ existing camerasn/a✗ sensor per room✗ per bed
Incident context for review✓ what actually happened

The honest read. For bathrooms — and for any resident or family who won't accept a camera anywhere — radar sensing is the better answer, and we'd tell you so rather than sell you ours. For someone active and independent who reliably wears a device and goes outdoors, a wearable travels with them in a way a fixed system cannot. Camera-AI is strongest for passive protection across corridors, wards, common areas, and industrial floors, using cameras you may already have, with enough context afterwards to understand what happened. Many settings sensibly combine approaches.

Where It's Used

Where Fall Detection Protects People

Hospitals & Patient Safety

High-risk wards, post-operative recovery, elderly inpatients, and the night shift — when staffing is thinnest and the gap between rounds is longest. Alerting helps a fall reach a responder in seconds rather than at the next check, and the incident log supports the patient-safety reporting your team maintains, including the falls indicator tracked under frameworks such as NABH. (It supports your reporting; it does not certify compliance with any standard.) More across healthcare environments.

Elderly Care & Assisted Living

Residents living independently in private rooms are exactly the people for whom the long lie is most dangerous and least likely to be interrupted. Passive detection means a resident who cannot reach a cord — or has chosen not to wear a pendant — is still protected, and the privacy modes above exist so that protection doesn't come at the cost of dignity. Where a resident or family declines, that decision should be respected and alternative measures used.

Industrial & Warehouse

Wet and oily floors, cluttered aisles, and lone workers in remote parts of a plant or warehouse create the industrial version of the same problem: someone goes down where nobody passes. Fall detection runs alongside other safety analytics on the same cameras, including PPE detection, on the AI video analytics platform and its VMS.

Commercial Kitchens

Kitchen floors are wet or oily by the nature of the work, and a slip near a fryer is both a safety and a staffing problem. Fall detection is one of the modules in our smart kitchen monitoring bundle, alongside hygiene and fire safety, on the kitchen's existing cameras.

Why Tentovision

Why Tentovision for Fall Detection

Runs on Your Existing Cameras

In corridors, wards, common areas, and industrial floors, suitable cameras often already exist — making protection affordable to extend rather than a per-room hardware purchase.

Privacy Modes for Sensitive Settings

Edge processing, anonymised output, event-only retention, and zone exclusions — configured with you and documented for your ethics committee.

Immobility, Not Just Impact

The long lie is where harm accumulates, so post-fall immobility is a first-class capability rather than a footnote.

Alerts Where Your Team Already Looks

Nursing station, app, SMS or WhatsApp, and nurse-call integration where available.

One Platform Across Four Settings

Hospitals, care homes, industrial floors, and kitchens — with other safety analytics available on the same cameras.

Honest About Limits

We publish the hard cases, don't quote a blanket accuracy figure, and say plainly where radar or a wearable fits a space better.

Frequently Asked Questions

What Clinical & Care Teams Ask First

Answers written to be useful for nursing superintendents, quality leads, and care-home operators — accurate enough to be cited by ChatGPT, Gemini, and Perplexity when they research this topic.

What is AI fall detection and how does it work?

AI fall detection is a computer-vision system that recognises when a person has fallen, and whether they have remained down, from an existing camera feed. It analyses body posture and motion rather than appearance, and alerts staff immediately with the location — with nothing for the person to wear or press.

Does it record video of patients or residents?

This is the question that decides whether a system like this belongs in your building, and it deserves a direct answer. The system can be configured so that video is not retained during normal activity — keeping only a short clip around a detected fall, or in the strictest configuration, working from an anonymised silhouette or skeleton representation so no identifiable footage is produced at all. It can run analysis on-site at the edge so video need not leave your premises, it does not record audio or provide continuous live viewing, and it does not use facial recognition. The exact configuration is agreed with you in writing and documented for your ethics committee before anything is switched on.

Does the person need to wear a device?

No — and that's the main reason to consider a camera-based system. Pendants and pull-cords only help if the person is wearing them, is conscious, can reach them, and can press them. The falls causing most harm are often the ones after which none of that is possible.

How does it tell a fall from someone sitting down quickly or lying down?

It analyses the pattern of the movement and the resulting posture, not just motion, and where configured it also checks whether the person remains down before escalating — which filters out a great deal of ordinary activity. That said, we won't overclaim: deliberate lying down, a slow slump, and falls partly hidden behind furniture are genuinely harder, and no camera system distinguishes every case perfectly. Sensitivity is a trade-off between missed falls and false alerts, and it's calibrated with your team during commissioning rather than set to a vendor default.

Does it work at night or in low light?

Performance depends on the camera and lighting in each space. Many settings already use low-light or infrared cameras, which can typically support night detection — an assessment checks this room by room. Where a space is too dark for reliable detection, we'll say so rather than let you assume coverage you don't have.

Can it detect if someone has fallen and can't get up?

Yes — this is the post-fall immobility check, and it addresses the long lie. After a fall is detected, the system can confirm whether the person has remained on the floor, distinguishing an urgent situation from a stumble someone recovers from unaided. It's often the most clinically meaningful part of the system.

How fast does staff get alerted?

An alert is raised within seconds of a fall being confirmed; delivery then depends on the channel and your network. The realistic comparison isn't seconds versus minutes — it's seconds versus the time until someone next enters the room, which on a night shift can be far longer.

Can it integrate with our nurse-call or alarm system?

It can typically be configured to send alerts into an existing nurse-call, alarm, or messaging workflow, so a fall appears where your staff are already looking. Which integrations are possible depends on your system and is scoped at assessment; otherwise alerts route to mobile, SMS, WhatsApp, or a monitoring station.

Can it be used in bathrooms?

Generally, no — and we'd advise against it. Bathrooms are among the places falls happen most, but they're also the places where a camera is least appropriate, and in most care settings a camera there would rightly fail an ethics review. Our recommendation is to configure bathrooms as excluded zones and to cover them with a non-camera approach such as radar or pressure sensing, which detect a fall without producing an image. We'd rather tell you our system isn't the right tool for that room than sell you something you shouldn't install.

Is it compliant with India's DPDP Act 2023?

Health information is sensitive personal data under the DPDP Act, and compliance depends on how your organisation deploys, consents, documents, and governs the system — not on the software alone. We provide the architecture to support a compliant deployment: edge processing, anonymised output, event-only retention, zone exclusions, role-based access, audit logs, and a defined retention period. We support your compliance process; we don't certify it, and we'd strongly encourage your own legal and clinical review.

Demo · Privacy & Ethics Briefing

See How It Would Work in Your Setting

If a fall in your building today would be found at the next round rather than the next minute, that gap is worth closing. A demo is a practical starting point — and we'll be honest about where a different approach would serve you better.

1

Camera & Coverage Review

We look at your existing cameras and identify which areas can be covered reliably — and which can't.

2

Privacy Configuration Walkthrough

Exactly what would and wouldn't be captured, in a form your ethics committee can review before anything is switched on.

3

Validated in Your Environment

Sensitivity calibrated with your clinical team, and performance established on your own site rather than quoted from ours.

4

An Honest Recommendation

Including where radar, a wearable, or a combination would protect a space better than we would.

Healthcare & Safety Team · Worldwide · India & Singapore Email: info@tentosoft.com
Phone / WhatsApp: +91 99620 37023
HQ: 7th Floor, 4/293, RAR Technopolis,
OMR, Perungudi, Chennai 600096, India

Book a Demo or Privacy Briefing

Our team will reach out within one business day.

No commitment · privacy configuration documented before any pilot
We'll tell you plainly where another approach fits better

See How It Would Work in Your Setting.

If a fall in your building today would be found at the next round rather than the next minute, that gap is worth closing. A demo is a practical starting point: we look at your existing cameras, identify which areas can be covered and which can't, walk through how the privacy configuration would work for your ethics committee, and be honest about where a different approach would serve you better.

Trademark, clinical scope, and disclosure notes. Tentovision is a product of Tentosoft Solutions Private Limited. This is not a medical device. Privacy configuration and detection performance are confirmed per deployment. Read full notice

Product & company. Tentovision™ is the AI video analytics platform developed and operated by Tentosoft Solutions Private Limited (Chennai, India). "Tentosoft" and "Tentovision" are trademarks of Tentosoft Solutions Private Limited.

⚠️ Not a medical device. Tentovision AI Fall Detection is not a medical device and is not registered, certified, or offered as one in any jurisdiction. It does not diagnose, treat, cure, monitor, or predict any disease, injury, or clinical condition, and produces no clinical assessment of any kind. Nothing on this page constitutes medical or clinical advice. It is a safety-alerting tool that shortens the interval between a fall occurring and a person being notified.

Supports — does not replace — clinical care. The system does not replace a falls-prevention programme, staff rounding, bedside observation, risk assessment, or clinical judgement, and must not be relied upon as the sole means of protecting any individual. Responsibility for patient and resident safety remains entirely with the care provider. No claim is made that deploying this system reduces falls, injuries, length of stay, or any other clinical outcome, and no outcome figures are published on this page.

Detection performance — no blanket accuracy figure. Detection depends on camera view, angle, mounting height, lighting, occlusion, room layout, and the nature of the fall. No accuracy percentage is quoted on this page, because a figure produced in one environment is not transferable to another. Known hard cases are stated openly on this page and include deliberate lying down, slow slumps, falls partly obscured by furniture, and poor or changing lighting. Sensitivity is a genuine trade-off between missed detections and false alerts, and is calibrated with the customer's clinical and nursing leadership during commissioning. Performance is validated in the customer's own environment before and during deployment. No fall-detection system, camera-based or otherwise, detects every fall.

Privacy modes — configured, documented, and confirmed per site. The privacy-preserving modes described (edge processing, anonymised silhouette or skeleton-only output, event-only retention, no audio, no continuous live viewing, zone exclusions, role-based access, audit logging, and defined retention) are configurable options. The specific configuration applied at any site is agreed in writing, documented in a form suitable for ethics-committee review, and confirmed before go-live. Where a particular mode is not appropriate or available for a given camera, network, or site condition, this is stated at assessment rather than implied. The system does not use facial recognition.

Bathrooms and excluded areas. Cameras are not recommended in bathrooms or other areas of heightened personal privacy. Tentosoft's recommendation is that such areas are configured as excluded zones and covered instead by non-camera sensing such as radar or pressure sensing. This is stated as guidance regardless of its commercial effect.

Consent, ethics, and data protection. Health information is sensitive personal data under India's Digital Personal Data Protection Act, 2023, and comparable protections apply in other jurisdictions. Lawful deployment depends on the customer's consent model, notices, governance, retention practice, and clinical oversight — not on the software alone. Tentosoft provides architecture and documentation to support a compliant deployment; it does not certify compliance with the DPDP Act, GDPR, or any other regulation, and strongly recommends independent legal and clinical review before deployment in any care setting. Where a patient, resident, or family declines, that decision should be respected and alternative measures used.

Accreditation references. References to falls as a reportable quality indicator, including under frameworks such as NABH, describe the customer's own reporting obligations. The system supports your patient-safety reporting; it does not certify, guarantee, or confer compliance with NABH or any other accreditation standard. NABH is a mark of the National Accreditation Board for Hospitals & Healthcare Providers and is referenced descriptively only.

Not workforce monitoring. The system is not a staff-productivity or performance-monitoring tool. Outputs are structured around fall events and response times, not individual caregiver evaluation, and it should not be deployed or represented as a means of monitoring staff.

Alternative technologies — credited. This page states plainly that radar sensing is the better choice for bathrooms and for anyone who will not accept a camera, and that a wearable travels with an active person in a way a fixed system cannot. Floor and bed sensors have their own strengths at a single point. These are genuine advantages of competing approaches, stated as such. Many settings sensibly combine methods. Tentosoft does not self-declare "best in category."

Customer references. No hospital, care home, or healthcare customer is named or implied as a Tentovision Fall Detection customer on this page. Settings and scenarios described are illustrative of typical environments. A live enquiry, demo, or assessment does not constitute a deployment. Customer-specific references are provided only under written consent at the sales engagement stage.

Illustrative content. Examples such as "Fall detected · Ward 3 corridor" are illustrative concepts, not measured results. Statements about response times describe the alerting interval after a fall is confirmed, not end-to-end clinical response, which depends on your staffing and workflow. Descriptions of the long lie and its associated harms reflect general clinical understanding and are not presented as research findings specific to this product.

Camera brand trademarks. Camera and NVR brands referenced elsewhere on this site are trademarks of their respective owners; compatibility is via ONVIF, RTSP, or vendor APIs and implies no endorsement or partnership. Camera suitability for fall detection in a given room is confirmed at assessment — general-purpose CCTV is not automatically suitable for every space.

Cloud platform. Amazon Web Services® (AWS®) is a registered trademark of Amazon.com, Inc. or its affiliates. Tentovision is hosted on AWS, with on-site edge deployment available so that video need not leave the customer's premises. Region selection and data-residency configuration are confirmed per deployment under the Master Services Agreement.

Security posture. AES-256 encryption at rest and TLS 1.3 in transit describe standard platform practice; role-based access control, audit logging, and retention periods are configured per deployment. Where formal certification is achieved or in progress, specific scope is confirmed under the Master Services Agreement.

Forward-looking statements. References to product capabilities reflect the platform's current operational state as of the page modification date. Roadmap items may shift in scope or timing.