A Score-Level Fusion Benchmark Database for Biometric Authentication

Abstract

Fusing the scores of several biometric systems is a very promising approach to improve the overall system's accuracy. Despite many works in the literature, it is surprising that there is no coordinated effort in making a benchmark database available. It should be noted that fusion in this context consists not only of multimodal fusion, but also intramodal fusion, i.e., fusing systems using the same biometric modality but different features, or same features but using different classifiers. Building baseline systems from scratch often prevents researchers from putting more efforts in understanding the fusion problem. This paper describes a database of scores taken from experiments carried out on the XM2VTS face and speaker verification database. It then proposes several fusion protocols and provides some state-of-the-art tools to evaluate the fusion performance.

Four Machine Learning Algorithms for Biometrics Fusion: A Comparative Study

“…The developed fusion schemes were tested on the publicly available XM2VTS face and speech database [13,14]. The XM2VTS database contains facial and speech data from 295 subjects, recorded during four sessions taken at onemonth intervals.…”

Classifier Combining Rules Under Independence Assumptions

“…It is also interesting to point out that this special case is a good explanation to one conclusion drawn in the previous work of Tax et al [9], which states that, in a two-class problem, the sum rule and the product rule achieve comparable performance [9]. Another assumption (15) involved in the sum rule is so strong that it would be violated in many applications. However, this independent assumption is actually a sufficient condition for the sum rule in our explanation.…”

“…This is the same sum rule that has been widely used in the multiple classifier system field. Compared to the product rule, two more independent assumptions, (8) and (15), are involved in the sum rule. To further understand these assumptions, some comments about these two rules are described as follows.…”

“…Our experimental data set 1 is presented by Poh and Bengio [15] to encourage researchers to focus on the problem of biometric authentication score-level fusion. The scores are taken from the XM2VTS database, which contains video and speech data from 295 subjects.…”

“…There exist two configurations called Lausanne Protocol I (LP1) and Protocol II (LP2) in the dataset with different partitioning approaches of the training and development sets. In both configurations, the test set is the same [15]. In our experiment, we pick the LP1 for our experimental study where eight different classifiers are available.…”