An automatic ear recognition approach

In ear recognition applications, noise and partial occlusion on the ear image are unavoidable. These two problems will severely degrade system performance. This paper proposes an automatic ear recognition system to deal with these two problems. The system mainly includes two stages: ear detection, ear feature extraction and classification. In stage one, an improved Adaboost algorithm is applied to realize real-time ear detection and tracking. In stage two, we propose a sparse representation based ear recognition approach. A test ear image to be identified can be represented as the sparse linear combination of the training images plus the sparse err produced by noise or partial occlusion on the test ear image. The experimental results on USTB ear database show the robustness of this sparse representation based ear recognition system, especially when the test image is corrupted by noise or partial occlusion.