Error-rate-based fusion of biometric experts

User verification using biometrics is not completely reliable due to false acceptances and false rejections. By employing multiple biometric experts and fusing their output, the reliability of the verification process can be improved. A common approach is score-level fusion, in which the match scores generated by each experts are fused into a combined score, which is then used to decide whether a claimant is genuine. Since different experts may have incompatible score distributions, score-level fusion requires careful considerations. We propose a novel solution to this problem with an error-rate-based fusion strategy, which relies on the false acceptance and false rejection rates of each individual expert to compute the overall confidence of the user verification. We demonstrate that our fusion strategy outperforms the sum rule score combination scheme and that it is robust to changes in the quality of the biometric samples.