NCJ Number
222343
Journal
Journal of Forensic Science Volume: 53 Issue: 2 Dated: March 2008 Pages: 325-330
Date Published
March 2008
Length
6 pages
Annotation
This study tested the repeatability and reproducibility of earprint acquisition according to "best practices," with attention to the importance of studying operator effects, operator training, and not collecting "matching" reference material at the same time.
Abstract
"Repeatability" refers to whether earprints from the same ear, taken repeatedly under the same circumstances by the same operator, are sufficiently similar. Study findings suggest that different operators may acquire earprints of differing quality, with equal error rates of the matching system ranging from 9 percent to 19 percent. Further, "matching" earprints are more alike when taken in a consecutive row than when taken on separate occasions. This shows the importance of studying operator effects, operator training, and not collecting "matching" reference material on the same occasion. The study, called the Forensic Ear Identification Project, collected a database of 1,229 earprint donors between 2002 and 2005. Three prints per ear were collected according to "best practices" for earprint acquisition. This report presents detailed descriptions of feature extraction and numerical evaluation. The countries participating in the research (Italy, the Netherlands, and the United Kingdom) each provided two investigators for the collection of earprints. The analysis focused on the country in which the earprint was taken, the donor of the earprint, the side (left or right) of the earprint, the operator (country of origin and which investigator in each country), and the number of the run (one up to four) in which the earprint was taken. 7 tables, 7 figures, and 7 references