Abstract
This paper presents three topics related to the goal of head pose classification: (1) improvement of the conventional state-of-the-art face detector; (2) implementation of head-pose classifier; (3) robust and fast eye-tracking. Proposed face detector is learned by replacing each training image by multiple binarized ones using multi-thresholds and detects face(s) in a binarized image rather than usual grayscale one. Implemented head-pose classifier reconstruct input image using principal components and best pose-class whose reconstruct matches the input most is selected. Robust and fast eye-tracking uses four templates of eye-corners and searches neighboring areas with highest score of normalized cross correlation.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
