Skip to Main Content
In this paper we present a feature partitioning approach to subspace classification. The proposed method computes subspaces using feature partitioning approach, where each pattern is divided into sub-patterns and extract features locally from sub- patterns and combines them to compute global subspace. We prove that the proposed approach consumes significantly less time in comparison to traditional PCA based subspace methods. The superiority of proposed approach can be understood from the experimental results of feature partitioning approach to principal component analysis over traditional principal component analysis.