查看文章

A uni-biometric system suffers from unbalanced accuracy because of image quality, features extraction weakness, matching algorithm and limited degrees of freedom. This can be overcome by using multiple evidences of the same identity (Multi-biometrics fusion). In a previous work, we proposed new fusion functions based on arithmetic operators and search the best ones using Genetic Programming on the XM2VTS score database. The objective function is based on the Half Total Error Rate (HTER) (a threshold dependent metrics), from the Expected Performance Curve (EPC), of fused matching scores. In this paper, we select ten functions from the generated ones and apply them on matching scores of different biometric systems, which are provided by the bio-secure database. This database provide 24 streams that we use to generate 1000 multi-biometric combinations that we, then, use to conduct our …