The architecture is composed of five parallel template detectors, a Radial Basis Function Network and two Kalman filters. The template detectors are five view-specific images of the head ranging ...