Using Density based Score Fusion for Multimodal Identification Systems under the Missing Data Scenario

While biometric fusion is a well-studied problem, most of fusion schemes cannot account for missing data (incomplete score lists), that is commonly encountered in large-scale multimodal identification systems. In this paper, we present a new approach, where RIBG (Robust Imputation Based on Group method of data handling) is used for handling the missing data. Since this scheme can be followed by a standard fusion approach designed for complete data, we propose a density based score fusion to achieve optimal performance in biometric systems. The rank-1 recognition rates of the proposed approach were 95.02% on the NIST-Multimodal database, 76.23% on NIST-Face database and 82.24% on NIST-Fingerprint database, even when the missing rate is set to 25%, which is higher than traditional approaches such as majority voting.