About ElderGuard
What Inspired Us
As the elderly population grows, the risk of fall-related injuries in nursing homes and care facilities continues to rise. Falls are one of the leading causes of injury and hospitalization among the elderly, and many of these incidents go unnoticed due to limited staffing and constant monitoring challenges. Inspired by this problem, we created ElderGuard, a real-time fall detection system designed to notify caregivers when someone is in a potentially dangerous situation. Our goal is to prevent falls from leading to serious harm by providing caregivers with an easy-to-use system that alerts them instantly.
What We Learned
Through this project, we deepened our understanding of computer vision and pose detection technologies, especially how they can be applied in real-world settings. We explored tools like MediaPipe for pose detection and OpenCV for real-time video processing, learning how to integrate them into a responsive web-based system using Flask. Additionally, we learned how to optimize our detection algorithm for reliability while keeping the user interface simple for non-technical caregivers.
How We Built ElderGuard
ElderGuard was built using a combination of powerful, open-source tools:
- Flask for building the backend and serving the web application.
- OpenCV for processing the video feed and detecting body positions.
- MediaPipe for real-time pose detection to determine if a person is lying down.
- Socket.IO for real-time communication between the server and client, ensuring immediate feedback in the interface.
- HTML/CSS/JavaScript for building a clean and user-friendly front end, allowing caregivers to easily monitor and interact with the system.
We also implemented real-time sound alerts and event logging to ensure that caregivers are notified immediately and can track historical events for later review.
Challenges We Faced
One of the main challenges was ensuring accurate pose detection even in environments where camera angles or lighting conditions may vary. We worked on fine-tuning the detection thresholds to balance sensitivity and reliability, minimizing false positives while still capturing critical events.
Another challenge was designing an interface that would be intuitive and efficient for caregivers, many of whom may not have technical backgrounds. We had to ensure the system was both easy to use and informative, allowing quick responses without overwhelming the user with unnecessary details.
Lastly, managing real-time performance was critical as we needed the system to respond instantly to potential falls while handling video stream processing in a lightweight manner. We addressed this by optimizing the communication between the client and server.
What's Next for ElderGuard
While ElderGuard currently provides a robust solution for detecting falls, we believe there is much more potential to explore. Future improvements could include:
- Machine learning integration for more advanced behavior prediction, allowing the system to detect potential falls before they happen.
- Mobile notifications via SMS or email to alert caregivers even if they are not physically near the monitoring station.
- Multi-camera support for monitoring larger areas with multiple individuals.
Log in or sign up for Devpost to join the conversation.