CLPEX Chat Board

Did anyone listen to the Forensics@NIST conference last week? There was a lot on likelihood ratios (L/R) and how to go about developing them?

I found this talk (Start at 1:13:00) by Elham Tabassi the most interesting with a few points of discussion:


The distilled takeaways are:

The factors feeding the priors are:

• # of minutiae
• clarity
• pattern type
• quality
• age or gender

The point at which L/R's are most needed (complex latents), they become the least powerful because of the factors that feed the priors.

Is anyone tracking in their LIMS when they encounter a complex print?