Two Dimensional PalmPhasor Enhanced by Multi-orientation Score Level Fusion

The security and protection of biometric template has been the bottleneck of its applications due to permanent appearance of biometrics. Texture codes are important approaches for palmprint recognition; unfortunately, there is no ideal cancelable scheme for palmprint coding until now. We propose a novel cancelable palmprint template, called “PalmPhasor”, which is a set of binary code fusing the user-specific tokenised pseudo-random number (PRN) and multi-orientation texture features of PalmCodes. PalmPhasor is extended from one dimension to two dimensions to reduce computational complexity. Two dimensional (2D) PalmPhasor in row orientation has better performance than one dimensional (1D) PalmPhasor. Furthermore, 2D PalmPhasor fuses multi-orientation texture features in score level to enhance recognition performance. Besides, the number of orientations of texture features for fusion is also discussed in this paper. The experimental results on PolyU palmprint database show the feasibility and efficiency of 2D PalmPhasor enhanced by multi-orientation score level fusion.