Kainat043/Posture-and-Ergonomics-Monitor-For-Meeting

🎥 AI Meeting Posture & Presence Monitor

An AI-assisted browser-based application designed specifically for meeting posture and presence monitoring.

The application uses real-time computer vision to estimate:

• —Meeting-ready posture
• —Good meeting posture
• —Poor meeting posture
• —Looking down / possible screen focus
• —Left and right leaning
• —Away / resting position
• —Camera distance
• —Person presence

The application runs directly in the browser using MediaPipe.

An optional Hugging Face Inference API integration can generate natural-language meeting guidance.

Features

🎥 Live Meeting Monitoring

• —Real-time webcam monitoring
• —Live MediaPipe Pose detection
• —MediaPipe Face Detection
• —Real-time skeleton overlay
• —FPS monitoring
• —Camera status indicator
• —Detection stability indicator

Meeting-Focused Detection

The application is intentionally designed around meeting scenarios.

Meeting-Ready Posture

The person appears to have a relatively suitable upright upper-body position.

Good Meeting Posture

Head, upper body and shoulder alignment remain within stricter thresholds.

Poor Meeting Posture

The system detects excessive deviation in one or more posture measurements.

Looking Down / Screen Focus

The head appears significantly downward.

This can indicate:

• —Looking at a laptop
• —Looking at a phone
• —Looking down at notes
• —Other downward head positions

The system cannot determine the exact reason.

Leaning Left / Right

The upper body appears shifted toward one side.

Away / Resting

A sustained steep head or torso position is detected.

This state is intentionally called:

Away / Resting

rather than "Sleeping Pose", because computer vision cannot reliably confirm that a person is sleeping.

Camera Too Close

The face occupies a large portion of the camera frame.

Person Not Detected

Reliable upper-body landmarks are not visible.

Meeting Posture Signals

The technical measurements are presented using user-friendly meeting terminology.

These values are estimated from MediaPipe Pose landmarks.

Meeting Posture Score

The application calculates a session-level percentage based on the proportion of monitored frames classified as acceptable meeting posture.

The score is an application-specific indicator.

It is not a medical measurement or clinical assessment.

Session Analytics

The dashboard provides:

• —Session Time
• —Meeting-Ready Time
• —Away / Resting Time
• —Posture Alerts
• —Person Absent Events
• —Camera Distance Alerts
• —Meeting Posture Score
• —Detection History

Technologies

• —HTML5
• —CSS3
• —JavaScript
• —MediaPipe Pose
• —MediaPipe Face Detection
• —Hugging Face Inference API
• —Hugging Face Spaces Static SDK

How MediaPipe Works

MediaPipe Pose detects human body landmarks from webcam frames.

Important landmarks include:

• —Nose
• —Left Shoulder
• —Right Shoulder
• —Left Hip
• —Right Hip

The application uses these landmarks to estimate:

Neck Alignment

Angle between the shoulder center and nose.

Sitting Alignment

Angle between the hip center and shoulder center.

Shoulder Balance

Difference in shoulder orientation.

Head Position

Horizontal displacement of the nose relative to the shoulder center.

The measurements are then passed through rule-based classification logic to determine the current meeting state.

Architecture

Webcam
   |
   v
MediaPipe Pose
   |
   +--------------------+
   |                    |
   v                    v
Body Landmarks     Face Detection
   |                    |
   +---------+----------+
             |
             v
     Posture Calculations
             |
             v
      Rule-Based State
        Classification
             |
      +------+------+
      |             |
      v             v
 Dashboard      Event History
      |
      v
Optional Hugging Face AI
      |
      v
Meeting Guidance