Multiple Classifier Fusion for Biometric Authentication

Abstract

Individual biometric modalities are continuously developed to improve their performance by sensor, system and algorithmic improvements. However, a very attractive alternative is to gain enhanced performance and robustness of biometric systems by combining multiple biometric experts. Recent research has demonstrated that both, the fusion of intra-modal experts as well as multi-modal biometrics impact beneficially on the system performance. In the former case the benefits derive from pooling the opinions of individual intra-modal experts. In the latter, complementary biometric information is brought to bear on the personal identity authentication problem. The issues involved in multiple biometric expert fusion and its potential will be discussed and illustrated on the problem of combining face and voice based identification.