Biometric Authentication: A Machine Learning Approach

10.4. Multisample Fusion

As discussed in the previous sections, the performance of biometric authentication systems can be improved by the fusion of multiple channels. For example, in Wark and Sridharan [371] and Jourlin et al. [169], the scores from a lip recognizer were fused with those from a speaker recognizer, and in Kittler et al. [182], a face classifier was combined with a voice classifier using a variety of combination rules. These types of systems, however, require multiple sensors, which tend to increase system cost and require extra cooperation from users (e.g., users may need to present their face and utter a sentence to support their claim). Although this requirement can be alleviated by fusing different speech features from the ...