Information-theoretical comparison of evidence evaluation methods for score-based biometric systems
Year: 2008
Type: conference-paper
Abstract: Biometric systems are a powerful tool in many forensic disciplines in order to aid scientists to evaluate the weight of the evidence. However, uprising requirements of admissibility in forensic science demand scientific methods in order to test the accuracy of the forensic evidence evaluation process. In this work we analyze and compare several evidence analysis methods for score-based biometric systems. For all of them, the score given by the system is transformed into a likelihood ratio ( ) which expresses the weight of the evidence. The accuracy of each computation method will be assessed by classical Tippett plots- We also propose measuring accuracy in terms of average information given by the evidence evaluation process, by means of Empirical Cross-Entropy ( ) plots. Preliminary results are presented using a voice biometric system and the NIST SRE 2006 experimental protocol. 1.
Source: Hispana
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3102998238
