NeoFace: Development of face detection and recognition engine

These years, great expectations are being held of biometrics authentication with faces, in view of increasing vicious crimes and terrorist threats. This paper describes a new face detection and recognition engine NeoFace developed at NEC from a technical point of view. For face detection, eye candidate detection and face determination by generalized learning vector quantization are combined to achieve high speed and high performance. For face recognition, two key technologies, the perturbation space method and adaptive regional blend matching, have been developed to reduce the adverse effects of pose changes, facial expression changes, wearing glasses and so on.