CTIT University of Twente
Research Business & Innovation About CTIT Research Calls Looking for a job? Intranet

Funding for facial recognition

When comparing a facial image from a crime scene with a police photograph, forensic experts pay attention to morphologic-anthropologic features, following a prescribed protocol. The support for the hypothesis that crime-scene and reference image originate from one individual is expressed in qualitative terms, such as “no support”, “limited support”, “moderate support”, “strong support”, and “very strong support”. Presently a strong effort is being made to replace such a qualitative description by a likelihood ratio, which is a quantitative measure of evidential value that can support the judge in court to make an objective decision.

Biometric face recognition is an automated process, for example used in access control. It does not use morphologic-anthropologic features, but extracts and compares abstract features, resulting in a quantitative similarity score. If this is above a predetermined threshold, the compared faces are accepted as resulting from the same individual.

 This proposal combines the fields of forensic face comparison and biometric face recognition in order to develop a (partially) automated system for forensic facial comparison that quantifies the evidential value as a likelihood ratio thus supporting the court to make an objective decision. This will be realized by:

  1. (Semi-)Automation of the Current Forensic Morphological-Anthropological Practice.
  2. Optimization of Biometric Face Recognition for Forensic Applications.
  3. Computation of Likelihood Ratios as Quantitative Evidential Value from the above 2 systems.