Real-Time Posture Detection & Feedback Using Machine Vision

This project at Human System Labs focuses on improving workplace safety in healthcare settings through real-time posture monitoring. I developed an OpenPose-based system to track body posture, identifying ergonomic risks that could lead to musculoskeletal injuries. By designing an Arduino Uno R4 WiFi device, I enabled real-time alerts for posture deviations, empowering workers to adjust their positions and reduce injury risks. This initiative showcases my expertise in machine vision and hardware integration to create practical, human-centered solutions for occupational health.

The setup uses a standard webcam for real time video input. Openpose was used to analyze the posture in real time. Whenever the posture was wrong, a signal was send to the arduino via WiFi. A combination of Python and C++ script was responsible for this seamless communication. The Arduino had 4 micro motors that were attached to the lower back of the participants. Upon receiving the signal for wrong posture, these motors would vibrate simultaneously thus providing the feedback.

A combination of eye metrics (pupil diameter, blink rate and fixations) along with task completion times, task performance accuracy and NASA-TLX was used to assess the effectiveness of real time feedback in a simulated workplace setting.

Key Contributions:

-Built an OpenPose-based system for real-time posture detection, tailored for healthcare environments.

-Engineered an Arduino Uno R4 WiFi device to deliver immediate feedback on posture risks.

-Analyzed posture data to provide insights into ergonomic improvements, enhancing worker safety and comfort.

-Collaborated with interdisciplinary teams to refine system design and ensure practical implementation.