An Effective Multi-Mode Iris Authentication System on a Microprocessor-FPGA Heterogeneous Platform With QC-LDPC Codes

Abstract

With the emergence and popularity of iris biometrics, there are increasing concerns regarding the feasibility of iris authentication systems and their corresponding variability reduction methods. The former issues are typically addressed by optimizing key factors, such as iris size, image quality and acquisition wavelength. As for the latter, introducing error correction codes to reduce intra-user variability in the enrolled identifiers becomes novelly promising. This paper proposes a conventional iris authentication system and a hardware-friendly QC-LDPC error correction code scheme on a microprocessor-FPGA platform. Different QC-LDPC codes in IEEE 802.16e were analyzed and selected. Suitable codes were applied, followed by the evaluation experiments. The proposed design achieves a competitive result with up to 0.20% EER and 0.50% ZeroFAR on the CASIA-IrisV4-Syn database. Cryptographic keys with lengths of up to 288 bits can also be generated and recovered. Such a device can be potentially used for applications such as an access control system in high-security areas, identity verification at the borders, biometric cryptography and related authentication scenarios.

Toward an Efficient Iris Recognition System on Embedded Devices

“…Grabowski and Napieralski [12] studied hardware efficiency for complex IR algorithms and remote camera sensors using Digital Signal Processors and FPGAs. Ma et al [15] implemented IR Quasi-cyclic low-density parity-check (QC-LDPC) error correction with FPGA. Also, Grabowsky et al [16] proposed an iris-on-the-move recognition system which does not require special cooperation with the user during biometric sample acquisition.…”

“…Now that a minimum iris radius of 45px has been established, we must correlate the iris radius with the distance to the camera. For this test, we captured four face images on the device at distances of 10,15,20,25,30,40,50, 60 and 70 cm from the camera. Then, the iris radii on the eight eyes contained in the four frames are manually measured, and the mean iris radius is scored for each distance value.…”