NMF方法在遮挡人耳识别中的应用

Applying NMF to Occlusion Ear Recognition

遮挡一直是人耳识别的典型难题,严重阻碍该生物特征识别技术的快速发展.非负矩阵因子NMF(non-negative matrix factorization)是一种线性子空间特征提取方法,近年来在识别中获得了比较成功的应用.将NMF的两种方法 NMFSE(NMF square error)和NMFDIV(NMF with divergence)以及在此基础上进行改进的LNMF(localNMF)、NNSC(non-negative sparse coding)、SNMF(sparse NMF)和NMFSC(NMF with sparseness constraints)等方法应用于遮挡情况下的人耳识别,并通过实验验证了NMF方法在遮挡条件下人耳识别的可行性与有效性.

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.