Forward Head Posture Alert - V3

V3

• Tkinter interface to adjust the sensitivity and duration of slouching after which you'll be notified
• posture_settings.json file to store user preferences
• Taskbar icon
• Changes to reduce computations :
  • skip every N frames.
  • reduced sound duration.
  • lower resolution for video feed.
  • pause detection process if no user is visible.

Overview

This project aims to detect a user's posture using a webcam and alert them if they have bad posture (forward head posture) that could lead to neck pain or other discomfort.

NOTE : Face and shoulders should be visible.

The application uses :

• Mediapipe for pose detection.
• OpenCV for face detection and visualizations.
• winsound and threading to play a sound (in a loop).

When the user's posture is detected as bad, the application plays an alert sound to prompt them to correct their posture.

The program captures webcam feed and continuously monitors the user's pose. It uses the position of the ears and shoulders to determine if the posture is good or bad. If the user's posture is detected as bad, the program plays an alert sound. Once the posture is corrected, the sound stops.

Current Features

• Posture Detection: The program analyzes the user's posture in real-time using Mediapipe Pose.
• Face Detection: Haar Cascade classifiers are used to detect faces (frontal and profile) to estimate ear positions when the pose landmarks are not detected.
• Alert System: An alert sound is played when the user has bad posture, and the sound stops when good posture is detected.
• Real-time Feedback: Posture status (good or bad) is displayed on the screen, along with colored visual indicators.
• Customizable Sensitivity: Users can adjust the sensitivity threshold and slouching duration through a Tkinter interface.
• Settings Persistence: Sensitivity threshold and slouching duration are saved in a posture_settings.json file and persist across sessions.

Limitations

• Accuracy of Ear Estimation: If the Mediapipe pose detection fails to detect ears, the program falls back to using the Haar Cascade face detection, which may not always give accurate ear positions, especially for side profiles.

Requirements

• Python 3.x
• mediapipe library for pose detection
• opencv-python library for computer vision tasks
• threading and winsound libraries for sound control
• tkinter for GUI interface

Install the required dependencies

To install the required dependencies, run:

pip install mediapipe==0.10.14 opencv-python tkinter