From seni Tue Nov 21 12:09 EST 1995
Date: Tue, 21 Nov 1995 12:08:59 -0500
From: Giovanni Seni <seni>
To: milun@cs.Buffalo.EDU
Subject: abstract for TR 95-38
Content-Type: text
Content-Length: 2445

A critical feature of any computer system is its interface with the
user. This has led to the development of user interface technologies
such as mouse, touchscreen and pen-based input devices. Since
handwriting is one of the most familiar communication media, pen-based
interfaces combined with automatic handwriting recognition offers a
very easy and natural input method. Pen-based interfaces are also
essential in mobile computing because they are scalable. Recent
advances in pen-based hardware and wireless communication have been
influential factors in the renewed interest in on-line recognition
systems.  

On-line handwriting recognition is fundamentally a pattern
classification task; the objective is to take an input pattern, the
handwritten signal collected on-line via a digitizing device, and
classify it as one of a pre-specified set of words (i.e., the system's
lexicon). Because exact recognition is very difficult, a lexicon is
used to constrain the recognition output to a known vocabulary. 
Lexicon size affects recognition performance because the larger the
lexicon, the larger the number of words that can be confused. Most of
the research efforts in this area have been devoted to the recognition
of isolated characters, or run-on hand-printed words. A smaller number
of recognition systems have been devised for cursive words, a
difficult task due to the presence of the letter segmentation problem
(partitioning the word into letters), and large variation at the
letter level. Most existing systems restrict the working dictionary
sizes to less than a few thousand words.   

This research focused on the problem of cursive word recognition. In
particular, I investigated the issues of how to efficiently deal with
large lexicon sizes, the role of dynamic information over traditional
feature-analysis models in the recognition process, the incorporation
of letter context and avoidance of error-prone segmentation of the
script by means of an integrated segmentation and recognition
approach, and the use of domain information in the postprocessing 
stage. These ideas were used to good effect in a recognition system
that I developed; this system, operating on a 21,000-word lexicon ,
was able to correctly recognize 88.1% (top-10) and 98.6% (top-10) of
the writer-independent and writer-dependent test set words
respectively.

----------------------------------END----------------------------

Thank you Davin.

-gs