Event-Based Face Detection and Tracking Using the Dynamics of Eye Blinks.

Area of Science:

• Computer Vision
• Robotics
• Biomedical Engineering

Background:

• Traditional face detection methods struggle with dynamic scenes.
• Event-based cameras offer high temporal resolution, ideal for capturing rapid biological signals.
• Eye blinks provide a unique, stable temporal signature for human faces.

Purpose of the Study:

• To develop the first purely event-based face detection method.
• To utilize the temporal signature of eye blinks for robust face detection and tracking.
• To leverage event-based sensors for capturing subtle facial dynamics.

Main Methods:

• Utilizing event-based cameras for high-speed data acquisition.
• Developing a correlation-based approach to detect the temporal signature of eye blinks.
• Implementing a probabilistic framework for spatial face tracking, corrected by blink events.

Main Results:

• Demonstrated successful face detection and tracking using only event-based data.
• Validated the method's efficacy in various indoor and outdoor experimental conditions.
• Showcased the stability and uniqueness of eye blink temporal signatures for identification.

Conclusions:

• Event-based sensing offers a powerful new paradigm for face detection.
• Eye blink analysis provides a robust biometric signature for computer vision tasks.
• The developed method paves the way for new applications in human-computer interaction and surveillance.