Improving Biometric Device Interoperability by Likelihood Ratio-based Quality Dependent Score Normalization

As biometric technology is being deployed, it will be a common situation to have a template acquired by one biometric device to be used by another during testing. In our problem formulation, we assume that the device may not be known to the system but can be inferred by a set of automatically derived quality measures. The proposed method of device inference results in a novel device-specific quality-dependent score normalisation called the DS-norm. Experiments based on the BioSecure DS2 data set confirm the benefits and feasibility of the proposed approach.