This awardee has received supplemental funding. This award detail page includes information about both the original award and supplemental awards.
Description of original award (Fiscal Year 2004, $490,700)
The goal of this project is to develop algorithms for use in analyzing handwriting in questioned documents. In previous phases of this research the principal investigator, in collaboration with Forensic Document Examiners (FDEs), developed algorithms for a subset of the handwriting features used by FDEs in the examination of evidence. A software system (known as CEDAR-FOX) automatically extracts and analyzes handwriting features and can compare a handwriting sample to a database (writer identification) or compare two handwriting samples to each other (writer verification). This project will now focus on further optimization, refinement, and testing of eighteen algorithms for extracting handwriting features. The software structure will also be re-designed to avoid system crashes and improve user interfaces for better navigation. This work will continue to be performed in close collaboration with FDEs at the State and Federal levels. The deliverable will be a prototype software containing 18 tested and validated algorithms for use in extracting and analyzing handwriting features in questioned documents.
1. Improve the algorithms used to extract writing elements.
2. Re-design and optimize the software system.
3. Collaborate with forensic science laboratories to evaluate, test, and refine the software to increase its forensic utility.
This project will extend previously developed computational approaches to handwritten questioned document (QD) examination previously funded by NIJ (2004-IJ-CX-K050). The proposed work under this cooperative agreement will focus on (i) development of a nine-point scale ranging from identified as same to identified as different, so that the software results will correlate with the opinions of QD examiners; (ii) improving automatic writer verification performance by removing a feature independence assumption; (iii) expanding the signature verification module to allow for multiple signature types and providing tools for automatic signature extraction from complex backgrounds; and (iv) studying disguised writing to see if any patterns emerge using computational techniques. The proposed tasks will be done with the guidance and assistance of forensic QD examiners.