NCJ Number
186080
Journal
International Journal of Forensic Document Examiners Volume: 5 Dated: December/January 1999 Pages: 138-145
Date Published
1999
Length
8 pages
Annotation
This paper presents a classification system for block capital writing based on the guiding principle that letter types should be as distinctly different as possible.
Abstract
Prior studies on classification systems of handwriting have relied on human operators to discriminate the sample pool. This may result in inconsistencies in classifying the same characteristic by two operators, or the same operator may classify the same characteristic differently on different occasions. Classification based on pattern recognition and image processing eliminates human error. The development of the system presented here has been one of progressive simplification from an original that contained many more types for each letter to one containing fewer types, each covering a wider range of letter forms and hopefully more distinctly separated from other types of the same letter. Block capital writing was chosen for study because it seemed likely to be better suited to an approach in which letters are classified individually and also because the published work deals more fully with cursive writing. The distinguishing features used were number of pen strokes and nature of pen path, the latter being preferred when both were available. Statistics are presented for the frequency of occurrence of the letter types in a sample of 140 writers. The number of letter types is comparatively limited, and some of the types occur with high frequency in the sample population. Nevertheless, computer-run studies have shown both that the classification system separates writers when sufficient letters are considered and that it has the potential for use as a retrieval medium. 1 table, 2 figures, 7 references, and appended supplementary information