Scribd
Upload
9 views2 pages

Fall Detection Project Script

The Fall Detection System using MediaPipe aims to provide a real-time, vision-based solution to detect falls among elderly individuals and patients. Built with a tech stack including Python, MediaPipe, and OpenCV, the system achieves 90-92% detection accuracy and operates with less than 100 ms latency. The user-friendly frontend displays live video feeds and alerts, making it suitable for healthcare and safety applications.

Uploaded by

adityavikramcodemaster
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views2 pages

Fall Detection Project Script

The Fall Detection System using MediaPipe aims to provide a real-time, vision-based solution to detect falls among elderly individuals and patients. Built with a tech stack including Python, MediaPipe, and OpenCV, the system achieves 90-92% detection accuracy and operates with less than 100 ms latency. The user-friendly frontend displays live video feeds and alerts, making it suitable for healthcare and safety applications.

Uploaded by

adityavikramcodemaster
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Fall Detection System using MediaPipe –

Presentation Script

Introduction
Good [morning/afternoon], respected faculty and friends. Today, I will be presenting my project: Fall
Detection System using MediaPipe Pose Estimation. The motivation behind this project is the
serious risk of falls among elderly people and patients. A fall that is not immediately detected can
lead to severe injuries or even death. Hence, my aim was to create a real-time, vision-based fall
detection system that is accurate, cost-effective, non-invasive, and capable of running on a normal
laptop or webcam.

Tech Stack Used


The system was built using a modern computer vision and web-based tech stack: - Python 3.10+ –
Core implementation language. - MediaPipe (v0.10.21) – Pose estimation with 33 body landmarks.
- OpenCV (v4.11.0) – Video capture and processing. - NumPy (v2.3.0) – Mathematical operations. -
CVZone (v1.6.1) – Helper utilities for overlays. - Flask (v3.0.3) – Backend server and API. -
Frontend – HTML5, CSS3, JavaScript for dashboard.

Backend Implementation
The backend is the brain of the system. It captures video using OpenCV, extracts landmarks using
MediaPipe, and computes mathematical features: - Torso Angle (θ): θ = arctan((y_hip - y_shoulder)
/ (x_hip - x_shoulder)) - Aspect Ratio (AR): AR = Width / Height - Vertical Speed (v_y): v_y = (y_t -
y_(t-1)) / ∆t - Knee Angle (φ): φ = cos^-1((AB · CB)/|AB||CB|) Fall is confirmed only if multiple
conditions are met: θ > 65°, AR > 0.8, Hip center > 0.75 frame height, Head y > 0.8, v_y > 0.03. A
buffer of 8–12 frames ensures robustness. The backend Flask API streams processed video and
alerts to the frontend.

Frontend Implementation
The frontend web dashboard shows: - Live video feed with skeleton overlay and bounding boxes. -
Status bar: Green = Normal posture, Red flashing = Fall detected. - Metrics panel: Displays torso
angle, aspect ratio, and confidence. - Alerts: Visual + sound notifications. - Responsive design
works on PC and mobile browsers.

Key Outcomes
- Achieved 90–92% detection accuracy in tests. - Runs in real time (25–30 FPS) with <100 ms
latency. - Filters out false positives like bending or sitting. - Works on standard webcam, no special
hardware required.

Conclusion
The project demonstrates that MediaPipe-based pose estimation combined with mathematical
analysis can detect falls in real time with high accuracy. The backend performs AI logic, while the
frontend provides a clean, user-friendly interface. This system is suitable for healthcare, smart
homes, and workplace safety applications. Thank you!

You might also like