摘要
针对人耳识别问题,提出了一种改进的稀疏性受限的非负矩阵因子(NMFSC)方法,通过增加一个使系数矩阵尽可能正交的约束条件来定义原目标函数,给出求解该目标函数的迭代规则,并证明迭代规则的收敛性。同时对人耳进行子区域划分,根据聚类规则对各子区域选择适当的权值,实现局部相似度到整体相似度的最佳映射。实验结果表明了该算法的优越性。
Based on ear recognition, an improved NMFSC (Non-negative Matrix Factorization with Sparseness Constraints) method was proposed by imposing an additional constraint on the objective function of NMFSC, which could capture the semantic relations of coefficient matrix as orthogonal as possible. The interated rules to solve the objective function with the constraint were presented, and its convergence was proved. The suitable weights were chosen for sub-region based on class-cluster role, which could get the optimal mapping overall similarity from sub-region similarity. The experiment results show that the proposed method can obtain better performance.
出处
《计算机应用》
CSCD
北大核心
2006年第4期790-792,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60375002)
关键词
人耳识别
非负矩阵因子
稀疏性
ear recognition
Non-negative Matrix Faetorization(NMF)
sparseness