Multispectral image modalities can offer high accuracy performance to the biometric systems by improving the discrimination along the spectral dimension. However, its adoptions are usually challenged by low signal to noise ratio, inter band misalignment, high data volume and computational efficiency. This paper presents a fast multispectral palmprint biometric technique using partial least square regression model and score-level fusion, with which an embedded recognition and verification system is implemented for evaluation. Our experiments are conducted using the PolyU palmprint database, and the results demonstrate that the proposed method can achieve a higher accuracy and a lower running cost compared to the reference implementations dedicated to the real-time and/or embedded applications.
A robust multispectral palmprint matching algorithm and its evaluation for FPGA applications
