Abstract:
Human Computer Interaction via Eye Blinks is a robust, accurate, and low-cost
real-time solution that automatically detects and measures user's eye blinks. It is
easy-setup eyes detection and tracking system, enables the user to interact with the
computer soon after catch ing the user's eye blinks. In addition, the system only needs
a 5-l 0 second calibration procedure at the very first time of use. An eye detection
method based on Motion analysis realizes these advantages. The system is intended to
provide an alternate input modality to allow people with severe disabilities and to
detect driver's drowsiness. Voluntary long blinks are caught whi le invo luntary short
blinks are ignored. The location of eye is automatically determined through the user's
initial eye blinks. Subsequently, the eye is tracked by Sum of Squared Difference
across time and appearance changes are analyzed to classify the eye as either open or
closed at each frame.
The system has been tested on several individuals with different facial color,
illuminating conditions and varied head poses; it returned qu ite a satisfactory
performance in both speed and accuracy. The algorithm is highly cost effective
