Supporting fingerprint identification assessments using a skin stretch model - A preliminary study

To support fingerprint expert opinion, this research proposes an approach that combines subjective human analysis (as currently applied by fingerprint practitioners) with a statistical test of the result. This approach relies on the hypothesis that there are limits to the distortion caused by skin stretch. Such limits can be modelled by applying a multivariate normal probability density function to the distances and angle formed by a marked ridge characteristic and the two closest neighbouring minutiae. This study presents a model tested on 5 donors in total. The "expected range" of distortion in a within-source comparison using 10 minutiae was determined and compared to between-source comparisons. The expected range of log probability densities for within-source comparisons marked with 10 minutiae was determined to be from -33.4 to -60.0, with all between-source data falling outside this range, between -83 and -305. These results suggest that the proposed generated metric could be a powerful tool for the assessment of fingerprint expert opinion in operational casework.