Xiyu Wang, Hengjian Li
Abstract
In order to extract more palmprint features and achieve better recognition results, a revocable palmprint recognition method based on Orthogonal Index of Maximum and Minimum (OIOMM) is proposed in this paper. Firstly, the competitive code features of ROI (region of interest) are extracted. Then, the statistical histogram of palmprint competition code features is obtained by partitioning the features. The Gaussian random projection (GRP)-based IOM mapping is used to generate GRP matrix. Orthogonal GRP matrix is obtained by Schmidt orthogonalization. OIOMM hash converts realvalued biological eigenvectors into discrete index hash codes. Finally, the palmprint image is matched with Jaccard distance. The experiment is carried out in the palmprint database of Hong Kong Polytechnic University. When the random projection size is 200 and the revocable palmprint feature length is 500, the equal error rate is 0.90. This shows that the algorithm not only improves the security, but also maintains the classification effect.
Keywords: Palmprint recognition, Orthogonal Index of Maximum and Minimum, competitive code, revocable palmprint template
**View Full Paper in the Springer Source