Palm-dorsa vein recognition based on independent principle component analysis

The traditional Principal Component Analysis (PCA) can only remove first- and second-order correlation between various components of data. To solve this problem and gain the features sensitive to the high-order information, two different architectures of independent principle component analysis (ICA) for palm-dorsa vein recognition are discussed, ICA architecture I based on statistically independent basis images and ICA architecture II based on statistically independent coefficients. A new approach based on invariant features is utilized to find the region of interest (ROI) in the palm-dorsa vein images so as to increase the recognition accuracy and reliability. We compare ICA with PCA on an existing palm-dorsa vein set. Experimental results show that two ICA architectures perform better than PCA and ICA, architecture I performs well, but not as well as ICA architecture II. Furthermore, the squared prediction error (SPE) of ICA is much smaller than that of PCA. It is illustrated that ICA can describe image features more essentially.