🚀 Inspiration

Initially, we built SafeSign for ICU patients who cannot communicate easily.

But we realized this problem exists everywhere — post-surgery rooms, elderly homes, and even people living alone.

So we expanded it into a universal human monitoring system.


💡 What it does

SafeSign monitors a person in real time using AI and detects:

  • Distress (gesture, facial expression)
  • Activity (eye blink, movement)
  • Unresponsive state
  • Critical conditions (fall or long inactivity)

It then sends instant alerts using:

  • Browser notifications (demo)
  • Telegram alerts (implemented)
  • SMS (extendable via Twilio)

🌍 Real-world scenarios

  • ICU: patient tries to signal but cannot speak
  • Post-surgery: detecting when patient regains consciousness
  • Elderly care: fall detection when no one is around
  • Home: monitoring people living alone

⚙️ How we built it

  • OpenCV for real-time video processing
  • MediaPipe for face, hand, and movement detection
  • Python backend with Flask
  • Telegram Bot API for alerts
  • Simple web dashboard for monitoring

⚠️ Challenges

  • Handling real-time detection without lag
  • Reducing false alerts
  • Making system universal instead of ICU-specific
  • Designing simple but effective alert system

🧠 What we learned

  • AI should solve real-world problems, not just demos
  • Simplicity is important in emergency systems
  • End-to-end integration (AI + alerts + UI) matters

🔮 What's next

  • WhatsApp alerts
  • Edge deployment (CCTV / Raspberry Pi)
  • Multi-person monitoring
  • Improved accuracy using deep learning

Built With

Share this project:

Updates