Eye Blink Detection Project

Overview

This project implements a real-time Eye Blink Detection system using MediaPipe Face Mesh and OpenCV. It calculates the Eye Aspect Ratio (EAR) to detect blinks, providing a valuable tool for applications in user interaction, health monitoring, and safety.

Key Concept: Eye Aspect Ratio (EAR)

The Eye Aspect Ratio (EAR) is used to determine whether the eyes are open or closed in real-time.The Eye Aspect Ratio (EAR) is a ratio of distances between specific facial landmarks around the eyes. It is used to determine whether the eyes are open or closed.The formula is:

EAR Formula:

EAR = (d1 + d2) / (2 * d3) Where:

• d1 = Vertical distance between landmarks p2 and p6 (top and bottom of the left side of the eye)
• d2 = Vertical distance between landmarks p3 and p5 (top and bottom of the right side of the eye)
• d3 = Horizontal distance between landmarks p1 and p4 (outer corners of the eye)

Explanation of the Formula:

1. d1 and d2: These are the vertical distances between the landmarks on the upper and lower eyelids of the eye. They measure how open or closed the eye is.
   • As the eye closes, these distances decrease.
2. d3: This is the horizontal distance between the outer corners of the eye, which provides a stable reference point that remains relatively constant regardless of head orientation, distance from the camera, or face size.

• Numerator: The vertical distances between the upper and lower eyelid landmarks (p2-p6 and p3-p5). This measures how "open" the eye is.
• Denominator: The horizontal distance between the outer corners of the eye (p1-p4). This serves as a stable reference point, which remains relatively constant despite head tilt, distance from the camera, or face size.

Why Use Horizontal Distance (d3)?

The horizontal distance between the outer eye corners is used to ensure the EAR calculation is consistent across various scenarios:

• Stability: The horizontal distance (d3) provides a consistent baseline that does not change significantly with different head poses, tilts, or face sizes.
• Scale Invariance: Whether the person is close to or far from the camera, this horizontal distance helps maintain consistency in the EAR calculation.
• Pose Robustness: Slight head rotations or tilts have minimal impact on this horizontal distance, making the EAR more reliable in real-world scenarios.
• Consistency Across Individuals: The horizontal distance provides a normalized reference, making the EAR applicable to different face shapes and sizes.
• Scale Invariance: Whether the face is near or far from the camera, the EAR remains proportional because the horizontal distance is stable.

Blink Detection Logic:

• Open Eye: The vertical distances (numerator) are relatively larger compared to the horizontal distance, resulting in a higher EAR value.
• Closed Eye: As the eyelids close, the vertical distances reduce, and the EAR value decreases.

A typical threshold for detecting a blink is when the EAR falls below 0.25. If the EAR stays below this threshold for several consecutive frames, a blink is detected.

Features

• Real-time Eye Blink Detection: Utilizes MediaPipe Face Mesh for accurate facial landmark detection.
• EAR Calculation: Computes the Eye Aspect Ratio to detect blinks.
• Customizable Blink Threshold: Parameters can be adjusted to fine-tune blink detection.
• Results Logging: Records blink detection results, including EAR values and thresholds, in a text file.
• Visual Feedback: Displays detected blinks on the processed images.

Requirements

• Python 3.x
• OpenCV
• MediaPipe
• NumPy
• SciPy

Usage

1. Place your images in the D:/frames/person11 directory. Ensure the images are named numerically (e.g., 1.jpg, 2.jpg, etc.).
2. Run the script:

   python <script-name>.py
   

The processed results, including blink counts and EAR values, will be saved in blink_results_p11.txt. Press q to exit the display window while processing.