Robust palm print verification system based on evolution of kernel principal component analysis
Salwani Ibrahim, Haryati Binti Jaafar, Dzati Athiar Ramli · 2014
Palm print is an emerging type of biometric that attracts researchers in biometrics area. As compared to the other biometric traits such as face, fingerprint and iris, the image quality of a fingerprint is robust with more information can be employed even though it is in low resolution. A new approach in feature extraction called evolution of kernel principal component analysis (Evo-KPCA) was proposed to speed up the processing time in the extraction stage. It used a reduced set density estimate (RSDE) to define a weighted gram matrix. As a result, the Evo-KPCA only extracted the most relevant and important information from a dataset. A total of 2400 palm print images was collected from three types of android mobiles. An experimental evaluation showed that the Evo-KPCA performed well in term of processing and accuracy compared to the region of interest (ROI), principle component analysis (PCA) and kernel principal component (KPCA) with the Genuine Acceptance Rates (GAR) of more than 98% and shorter processing time of less than 0.5s.