Human identification and verification based on multimodal fusion of face and ear

This paper addresses the problem of multimodal fusion of frontal and profile face with ear for human identification and verification. A scale space local feature matching method is proposed to match both 2D texture and 3D shape index images, and various multimodal fusion strategies are investigated. Owning to use of the advantage of scale space local features, the proposed method is insensitive to scale variations, and is robust against noise and data incompleteness. The results of the experiment on the FRGC v2.0 and UND Collection F databases showed that the method's outstanding identification and verification accuracy was achieved, with the high robustness against expression variations because of the fusion of multimodal features.