作者: Dongliang Huang , Winston Li , Henry Leung
DOI:
关键词: Engineering 、 Sensor fusion 、 Biometrics 、 Data mining 、 Information fusion 、 Raw data 、 Fusion 、 Feature (machine learning) 、 Majority rule 、 Information system
摘要: Abstract : In this report, an overview of information fusion techniques for dependent and independent sources, specifically biometric applications, is provided. The architecture presented both sources addressing in detail the various at four different levels namely: raw data or signal level, feature decision level multi-level integrated fusion. Furthermore, report addresses question whether can be fused to provide multi-modal system with enhanced performance. shows that even when are independent, performance a better than based on single source. measured terms total false accept rate (FAR) rejection (FRR). conditions achieving improved using AND, OK majority voting derived theoretically confirmed through computer simulations.