Physio_Vision

• 

  Quick Illustration of the Functionality

What problem does it solve? (Value Proposition) With the aging global population, continuous in-home monitoring is critical, yet dedicated medical alert systems are often expensive and stigmatizing. Physio-Vision leverages devices that users already own and are comfortable with—their Amazon Fire TV and its camera. By acting as a passive, privacy-first monitor, it instantly detects falls, triggers local alarms, and logs critical health data, providing peace of mind for both seniors and their families.

How we built it (Technical Architecture) Developed the prototype using Python and OpenCV to simulate the Fire TV camera environment. To ensure the application remains ultra-lightweight and capable of running smoothly on edge devices (like Smart TVs), bypassed heavy machine learning models. Instead, implemented a highly optimized background subtraction and contour analysis algorithm. By calculating bounding box aspect ratios dynamically, the system distinguishes between standing and falling states with near-zero latency.

Challenges we ran into Initial approach involved using heavy pose-estimation models (like MediaPipe). However, quickly realized that integrating these into a constrained environment could lead to lag and high compute costs, contradicting the goal of an accessible TV app. The biggest challenge was pivoting under a strict time constraint to a mathematical, contour-based heuristic model. Tuning the aspect-ratio thresholds to accurately detect falls without triggering false positives during normal activities required rapid iteration and testing.

Built With

• computer-vision
• csv
• grid
• opencv
• python