A Novel Serial Multimodal Biometrics Framework Based on Semisupervised Learning Techniques

摘要: We propose in this paper a novel framework for serial multimodal biometric systems based on semisupervised learning techniques. The proposed addresses the inherent issues of user inconvenience and system inefficiency parallel systems. Further, it advances by promoting discriminating power weaker but more convenient trait(s) saving use stronger less whenever possible. This is contrast to other existing that suggest optimized orderings traits deployed parameterizations corresponding matchers ignore most important requirements common applications. In terms methodology, we techniques strengthen matcher(s) trait(s), utilizing coupling relationship between traits. A dimensionality reduction method dependence maximization achieve purpose. Experiments two prototype clearly demonstrate advantages methodology.