Muli-pose Ear Recognition Using Sparse
Yin Tian · Jisuanji fangzhen · 2014
In recent years, sparse representation is a hot area of research in the field of pattern recognition. A new and feasible scheme was investigated to solve the problem about limited viewing angles of images in ear recognition field. So the method for more accurate ear recognition using the sparse representation of HOG Descriptor was presented. The system was established by three stages: ear feature extraction, reducing the feature dimension and classification. HOG features of ear images were firstly extracted, then the dimensions of them were reduced by LDA algorithm and verification problems were solved by using the sparse representation. A test ear image to be identified could be represented as the sparse linear combination of the HOG features of training images. For pose variations of ear images, the comparative experimentresults indicate that the method has more accurate and stable ear recognition rate within certain range of feature number. And the method is robust for cases where rapid ear angle change exists.