Researchers have found a way to mathematically calculate the rarity of a fingerprint.Although fingerprints are unique to every individual, crime scene prints are usually incomplete patterns taken off doorknobs or glass.
Knowing the rarity of a partial print could be useful to forensic scientists who are trying to determine how valuable a fingerprint is as evidence, said Sargur Srihari, a computer scientist at the University at Buffalo who is leading the research.
... Dr. Srihari and his graduate student Chang Su say they have done the same for fingerprints [used the rarity of fingerprints to evaluate fingerprint evidence].
“It’s purely mathematical,” Dr. Srihari said. “We’re simply saying, ‘We just found something that is unusual, and that makes it an important piece of evidence.’ ”
To do the research, the scientists defined fingerprints as a series of points, composed of the endings of ridges and ridge bifurcations.
They then pulled from a database of 4,000 fingerprints kept on file at the National Institute of Standards and Technology and created a computer system that can read fingerprint patterns. Based on a print’s points, the system can mathematically determine its rarity.
...
The research was presented this week in Vancouver, British Columbia, at the annual Neural Information Processing Systems conference.
It's here: the law of evidence on Spindle Law. See also this post and this post.
1 comment:
So far the consensus among my mathematically- and statistically-sophisticated friends is that Sargur Srihari's approach will not pan out. However, my sources believe that mathematical and statistical analysis (and perhaps automated analysis?) is in principle possible and they recommend the work along these lines that Christophe Champod is doing. See his bio at https://applicationspub.unil.ch/interpub/noauth/php/Un/UnPers.php?PerNum=860967&LanCode=8
Post a Comment