Multi-pose face detection for silver robots

Abstract

This paper presents a method which can detect multi-pose faces for robot. To overcome a disadvantage of conventional single-view face detection. we compute hyperplane using Linear Discriminant Analysis (LDA). These hyperplane are used to separate two different classes finely. This method could reduce false positive rate on branching nodes significantly, as well as the number of unnecessary branches. Consequently, overall speed of the detector could be improved about 24 percent compared to the case without hyperplane partition.