The performance of the LR method was evaluated using different performance measures (Rates of Misleading Evidence, Cllr and EER) and their corresponding graphical representations: Tippett, ECE, and DET plots. The LR method used showed coherent behaviour: performance increased with the number of minutiae increasing from 5 to 10. It also showed somewhat incoherent behaviour and a small decrease in performance when moving from 10 to 11 minutiae.
This incoherent behaviour of the comparison algorithm's performance is believed to be caused by a switch of the method it Dacinostat uses when more than 10 minutiae are mainly due to the Archaebacteria core comparison algorithm used by the AFIS technology. The experimental example therefore reveals the importance of coherence in order to detect points of improvement in computer-assisted LR methods.
AcknowledgementsThis research was conducted in the scope of the BBfor2–European Commission Marie Curie Initial Training Network (FP7-PEOPLE-ITN-2008 under Grant Agreement 238803) at the Netherlands Forensic Institute, and in collaboration with the ATVS Biometric Recognition Group at the Universidad Autonoma de Madrid and the National Police Services Agency of the Netherlands.