LifeGuard AI

HEALTHCARE AND WELLNESS TRACK

Inspiration

We wanted to create an AI-powered solution that improves CPR quality and boosts confidence during emergencies. Real-time feedback can mean the difference between life and death, especially for untrained responders.

What It Does

LifeGuard AI uses MediaPipe Pose and motion analysis to detect chest compressions in real time from a camera feed. It validates rhythm, speed, and depth, then provides instant visual, audio, and metronome feedback to guide users through effective CPR cycles.

How We Built It

• Pose Estimation: Integrated MediaPipe for human landmark detection.
  
• Motion Analysis: Used pixel-based motion comparison to detect chest compressions.
  
• State Machine: Managed compression phases and tracked session cycles, including prompts for rescue breaths.
  
• User Interface: Delivered live stats, voice feedback, and rhythm tracking in real time.
  

Challenges

1. Ensuring accurate skeleton detection across diverse body types.
   
2. Precisely calculating chest coordinates using shoulder and hip landmarks for reliable motion focus.
   
3. Handling real-time video processing efficiently without latency on standard hardware.
   

Accomplishments

• Built a fully functional real-time CPR monitoring system.
  
• Achieved high accuracy in compression detection.
  
• Delivered immediate coaching cues, demonstrating how AI can make CPR training accessible to anyone, anywhere.
  

What We Learned

• How to combine pose estimation, motion tracking, and real-time feedback loops effectively.
  
• Techniques for tuning ML-based systems for dynamic human movement detection.
  

What's Next

• Integrate depth sensors or LiDAR for improved compression depth estimation.
  
• Add mobile app support for wider accessibility.
  
• Train a custom AI model for automated quality scoring and emergency guidance.
  

Built With

• canvas-api
• getusermedia
• google-gemini-ai
• mediapipe-pose
• react
• tailwind-css
• typescript
• vite
• web-audio-api
• web-speech-api