Since in the feature space the eigenvector is a linear combination of all the samples from the training sample set, the computational efficiency of KPCA-based feature extraction falls as the training sample set grows. In this paper, we propose a novel KPCA-based feature extraction method that...