15th International Conference on Pattern Recognition, ICPR-2000
573
INK-LINK
Adnan El-Nasan, George Nagy
Rensselaer Polytechnic Institute (RPI)
Troy, NY, 12180 USA
elnasan@ecse.rpi.edu, nagy@ecse.rpi.edu
Abstract
Wide acceptance of inexpensive writing tablets
with high functionality motivates the development of
individualized, adaptive on-line recognition of
cursive script. We demonstrate a lexical algorithm
based on bigram matches. The solution we propose is
to (i) Generate a match list by partial-word matching
against a reference list in the owner's script. (ii)
Identify each unknown word by eliminating, from a
large lexicon, every word that partially matches the
transcript of any word on the reference list that is not
on the match list, or that fails to match any word on
the match list. With perfect feature-level matching, a
surprisingly short reference list yields a high
recognition rate.
1. Introduction
Recent
research
in
Intelligent
Character
Recognition (ICR) has focused on off-line, multi-
writer applications such as bank-check interpretation
[3], postal sorting [9], and form reading [12]. These
applications make effective use of
restricted
vocabularies for various document components, of
inter-field redundancy, and of external databases
(legal and courtesy amount agreement; street, city,
state, zip-code;
item-name, item-number, price;
customer-name, customer-address; or entry, subtotal,
total). In some forms-reading applications, additional
constraints, like boxes, combs, or examples of
preferred writing styles, may be preprinted, possibly
in a drop-out color. Furthermore, in such large,
centralized applications an operating point with a
30% reject rate and a 0.1% field error rate may be
economically acceptable.
We are interested, in contrast, in decentralized,
individualized, unconstrained applications of on-line
recognition. The advent of slim, inexpensive tablets
with built-in processors will give new life, with new
requirements,