Human ear recognition based on block segmentation

A new human ear recognition approach based on block segmentation is proposed in this paper. In this method, an original ear image is partitioned into several smaller sub-images, then the sub-images are extracted by features, As a result, the lower dimension space features that can replace the original images are obtained. Finally the pattern classification can be implemented by the nearest neighbor classifier. To verify the effectiveness of the block segmentation approach, a various experiments are conducted based on four feature extraction methods. USTB human ear database is applied to test the algorithms. The experimental results indicate that the recognition rates are significantly improved. The recognition rate of the moment invariants based on block segmentation achieves 100% in the experiments. The statistics feature extraction is easy to actualize, and the computational speed of recognition is the fastest.