Algorithms for a novel touchless bimodal palm biometric system

Abstract

The paper introduces the combination of algorithms for possibly the first bimodal biometric system capable of touch-less capturing of two biometric parameters, palm veins and palm creases, synchronously with a single image sensor. The architecture of the proposed system is based on the Detection, Alignment and Recognition pipeline. The ROI detection and alignment stages are simplified with efficient combination of hardware (lighting sources) and software. A new feature descriptor, namely Histogram of Vectors is proposed in the recognition stage. Since the capturing of images requires special conditions, the database including images of 100 individuals and ground-truth data is introduced. The analysis of performance of the system utilizes the database leading to detailed understanding of the error propagation in the automatic recognition pipeline.