Abstract: Detecting and tracking human faces in a complex background is one of the most important multimodal human machine interfaces. This paper presents a human face detection and tracking system which is based on eigen subfaces. A standard human face template is established and a method of template matching is used as the first step. The concept of eigen subface is then proposed and a set of effective bases for human face’s organ space are designed. The eigen subface’s resolution measure and the strategy of human face detection and tracking are also presented are designed. Experimental results show that this system could detect and track human faces satisfactorily in various unconstrained backgrounds.