Invariant biometric code extraction
Hyung Gu Lee, Seung-In Noh, Kwanghyuk Bae, Kang-Ryoung Park, Jaihie Kim · 2005
We propose an invariant key generation method using an iris image. In conventional iris recognition systems, iris features are not consistent because most feature extraction techniques are sensitive to variations of the captured iris image data. We introduce an invariant binary feature which is defined as the iris key. Regardless of iris image variations, we can extract the invariant iris key in our experiments. Our method can be applied to various fields, especially for key protection and key management applications using biometric data; these include e-commerce, smart cards, and logging on to a personal computer or secure Web site. To generate the iris key, we propose a reference pattern which is designed as a lattice structured image to represent a bit pattern as the individual's identity. The reference pattern and iris image are linked into a filter. In the filter, iris texture is reflected according to the magnitude of the iris power spectrum in the frequency domain. There is a trade-off relationship between within-class tolerance and between-class discrimination, so that we can control these factors which are included in the filter. In experimental results, we were able to tune the system to a false rejection ratio (FRR) level of 1.0%.