Applying NMF to Occlusion Ear Recognition
Yu Wang · Journal of Zhengzhou University · 2013
Occlusion always was a difficult problem in the ear recognition,and hindered seriously the development of this biometric technology.NMF(non-negative matrix factorization) was a kind of feature extraction technique on linear subspace,and was successfully applied to face recognition in the last years.Two kinds of NMF methods including NMFSE(NMF square error) and NMFDIV(NMF with divergence),and some improved techniques such as LNMF(local NMF),NNSC(non-negative sparse coding),SNMF(sparse NMF) and NMFSC(NMF with sparseness constraints) were applied to occlusion ear recognition.The feasibility and effectiveness of these techniques were testified using experimental results in the occlusion condition.