Likelihood ratio estimation in forensic identification using similarity and rarity

Highlights：

• We formulate probability of identification in terms of likelihood ratio (LR).

• We factor LR into the product of two factors, based on distance and rarity.

• We describe methods for computing distance in forensics for different data types.

• We provide empirical evaluations of three methods of LR computation.

• Proposed method outperforms distance method with forensic data.