Search The Site
 
More options | Back issues
Home
News
Opinion & Forums
Careers
Multimedia
Chronicle/Gallup
Leadership Forum
Technology Forum
Resource Center
Campus Viewpoints
Services
/r

The Chronicle of Higher Education
Tuesday, June 11, 2002

Scientists Turn to Computers to Prove That Everyone's Handwriting Is Unique

By ANDREA L. FOSTER

Scientists at the State University of New York at Buffalo have turned to computers to verify that each person's handwriting is unique, according to a paper that will be published in the Journal of Forensic Sciences in July.

The paper could prompt more courts to accept handwriting analysis as evidence, says Sargur N. Srihari, who headed the study. Mr. Srihari, a computer-science professor, also directs the university's Center of Excellence for Document Analysis and Recognition.

Mr. Srihari recently testified about his research in an effort to help the government prosecute Michael Stefan Prime, of Seattle. A jury of the U.S. District Court in Seattle last month convicted Mr. Prime of conspiracy and using counterfeit money in connection with spurious sales on eBay.

Judges frequently do not allow testimony that identifies someone based on a visual analysis of handwriting because the analysis is not supported by scientifically rigorous criteria. In dismissing such evidence, judges often refer to a 1993 Supreme Court case, Daubert v. Merrell Dow Pharmaceuticals Inc. In that case, the court ruled that expert testimony must have valid scientific backing to be admitted as evidence.

Mr. Srihari says his ongoing research will give more credibility to the testimony of handwriting experts. They examine the spacing of words and the curve of letters, among other characteristics, to help determine whether a particular individual was the author of a handwritten text or fragment.

Mr. Srihari used artificial intelligence to help develop software that, he says, determines with 96 percent accuracy whether the same person wrote two different handwriting samples. The software applies pattern-recognition techniques to analyze 11 features that make up the overall structure of the writing, such as the layout of the document and spacing of each line, as well as 512 features of individual characters, such as individual stroke marks and the openings and closures of each character.

"We're trying to push the 96 percent to a higher accuracy," says Mr. Srihari.


Print this article
Easy-to-print version
 e-mail this article
E-mail this article




Headlines

Lilly Endowment to match up to $138-million in gifts to Indiana colleges

Goddard College plans to eliminate residential programs amid growing financial difficulties

Tel Aviv U. settles strike, but faces huge deficit

Medical school in Slovakia expected to abandon affirmative-action plan for Gypsy students

13 new chief executives announced

Scientists turn to computers to prove that everyone's handwriting is unique

Online education's drawbacks include misunderstood e-mail messages, panelists say


Copyright © 2002 by The Chronicle of Higher Education