How It Works

How Ilonai Works

Privacy-first fall detection using edge AI. All processing happens locally in your facilityβ€”video never goes to the cloud.


The Process: 4 Simple Steps

πŸ“Ή

1. Smart Cameras Monitor

Discreet cameras observe living spaces continuously

πŸ€–

2. AI Detects Falls

Edge AI identifies fall patterns in real-time locally

πŸ””

3. Staff Alerted

Notifications sent within 30-90 seconds

πŸ”’

4. Emergency Access Only

Video accessible only after confirmed fall


Step 1: Smart Camera Monitoring

Strategic Placement

πŸ›οΈ

Resident Rooms

Ceiling-mounted for full coverage, 1-2 cameras per room

🚿

Bathrooms

Privacy-conscious placement, no direct shower view

🍽️

Common Areas

Dining rooms, lounges, hallways

Camera Specifications

  • βœ“ 1080p HD resolution at 30 fps
  • βœ“ Night vision with infrared LEDs
  • βœ“ 4-6 weeks battery life + AC power option
  • βœ“ Discrete design similar to smoke detector
  • βœ“ No microphones β€” audio never recorded
  • βœ“ Status LED shows when active

Step 2: Edge AI Fall Detection

What is "Edge AI"?

Edge AI means artificial intelligence runs locally on a device in your facility, not in the cloud.

Ilonai’s Edge Device:

  • Hardware: Industrial-grade AI processor
  • Location: Secure closet in facility
  • Size: Compact (15cm Γ— 10cm Γ— 5cm)
  • Function: Real-time video processing

How AI Detects Falls (in milliseconds):

  1. Pose Estimation - Detects body keypoints and tracks position
  2. Motion Analysis - Measures velocity and sudden downward movement
  3. Context Understanding - Recognizes room layout and normal activities
  4. Fall Classification - Confidence score >85% triggers alert

What Happens to the Video?

βœ… Normal Operation (No Fall)

  • β€’ Video analyzed frame-by-frame
  • β€’ Only metadata extracted
  • β€’ Video immediately discarded
  • β€’ Never stored, never transmitted
  • β€’ All processing on edge device

⚠️ Fall Detected

  • β€’ 10 sec before + 30 sec after saved
  • β€’ Clip encrypted and stored locally
  • β€’ NOT uploaded to cloud
  • β€’ Alert sent to staff
  • β€’ Auto-deleted after 7 days

Step 3: Instant Staff Alerts

Multi-Channel Notification System

πŸ“±

SMS Alerts

Works even if internet is down (GSM)

πŸ“²

Mobile App

Shows location, time, resident info

πŸ’»

Dashboard

Real-time visual and audio alerts

Alert Escalation: If not acknowledged within 60 seconds, alert escalates to supervisor, then administrator


Step 4: Emergency Video Access

πŸ” Authentication Process

  1. 1. Staff logs in to dashboard or mobile app
  2. 2. Two-factor authentication required
  3. 3. View list of detected falls
  4. 4. Request access to specific incident
  5. 5. System verifies permissions
  6. 6. 40-second clip becomes available

πŸ“‹ Why Video Access Matters

Assess Severity
Did they hit their head? Are they conscious? Any visible injuries?

Determine Response
Send one nurse or full team? Call ambulance? What equipment needed?

Document Incident
For medical records, family communication, liability protection

Complete Audit Trail: Every access is logged with who, when, how long, and whyβ€”tamper-proof for compliance


System Architecture

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚         Care Facility (On-Premise)       β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚                                          β”‚
β”‚  πŸ“Ή Camera 1    πŸ“Ή Camera 2   πŸ“Ή Camera N β”‚
β”‚  (Room A)       (Room B)       (Hallway) β”‚
β”‚       β”‚             β”‚              β”‚     β”‚
β”‚       β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β”‚
β”‚                     β”‚                    β”‚
β”‚           β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β–Όβ”€β”€β”€β”€β”€β”€β”€β”€β”          β”‚
β”‚           β”‚  🧠 Edge AI Unit  β”‚          β”‚
β”‚           β”‚  NVIDIA Jetson   β”‚          β”‚
β”‚           β”‚                  β”‚          β”‚
β”‚           β”‚ β€’ Video Processing          β”‚
β”‚           β”‚ β€’ Fall Detection β”‚          β”‚
β”‚           β”‚ β€’ Encrypted Storage         β”‚
β”‚           β”‚ β€’ Alert Generation          β”‚
β”‚           β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”˜          β”‚
β”‚                     β”‚                    β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                      β”‚
                      β”‚ (Metadata only)
            β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
            β”‚                   β”‚
            β–Ό                   β–Ό
   πŸ“± Staff Devices      πŸ’» Admin Portal
   β€’ SMS Alerts          β€’ Dashboard
   β€’ Mobile App          β€’ Reports
   β€’ Notifications       β€’ Audit Logs

πŸ”’ Video stays in facility
βœ… GDPR compliant by design

Privacy Protections

βœ… What We DO

  • βœ“ Analyze video locally on facility premises
  • βœ“ Detect falls with >95% accuracy
  • βœ“ Alert staff within 30-90 seconds
  • βœ“ Emergency video access with authorization
  • βœ“ Encrypt all stored data
  • βœ“ Auto-delete footage after 7 days
  • βœ“ Log all access for accountability

❌ What We DON'T DO

  • βœ— Upload video to the cloud
  • βœ— Allow live monitoring by staff
  • βœ— Record audio
  • βœ— Use facial recognition
  • βœ— Share data with third parties
  • βœ— Use data for marketing
  • βœ— Allow remote access by Ilonai

Real-World Example

A Day with Ilonai: Mrs. Schmidt's Fall

πŸ•™ 10:30 AM - The Fall

Mrs. Schmidt (82, Room 204) trips on carpet edge while returning from dining room

T+0 sec: Camera captures fall

T+1 sec: AI confirms fall (94% confidence)

T+2 sec: Alert generated

T+4 sec: SMS sent to Nurse Anna

T+5 sec: Dashboard alert appears

T+8 sec: Mobile app notification

T+30 sec: Nurse Anna arrives

T+40 sec: Views video, confirms no head injury

βœ… Outcome

Bruised hip but no serious injury. Fast response prevented complications from prolonged floor time.

40 sec

Total response time

94%

Detection confidence

0

False alarms today


Technical FAQs

How much bandwidth does Ilonai use?

Very little β€” video stays local.

  • Normal operation: <1 MB/day per unit
  • During fall incident: 5-10 MB (if video viewed remotely)
  • Software updates: 50-100 MB/month

Total: ~200-300 MB/month per unit (10x less than one Netflix episode)

What if internet goes down?

Built-in redundancy ensures alerts are never lost:

  1. GSM Backup: Built-in cellular modem sends alerts via mobile network
  2. Local Network: Facility dashboard still receives alerts via local network
  3. Buffering: Unsent alerts stored and transmitted when connectivity restores
How does it work with pets or service animals?

AI is trained to detect human falls only. Animal movements are ignored due to:

  • Different body proportions (four legs vs two)
  • Different motion patterns
  • Different size and shape

Edge case: Very large dogs may initially trigger false alarms, but system learns after 1-2 corrections.

Can the AI be fooled?

Difficult but possible. However:

  • Dropping large objects may trigger false alarm (staff can mark and system learns)
  • Intentional slow falls may not detect (defeats the purpose)
  • Camera obstruction detected and alerts staff

Key point: Residents generally don't try to fool the systemβ€”they want the safety it provides.


Ready to Learn More?

See Ilonai in Action

Schedule a demo to see how our technology protects residents

πŸ“§ info@ilonai.ch β€’ πŸ“ž +41 76 404 05 21