Robotics Reference
In-Depth Information
Once a character has been separated out by the system, so that the
software is not looking at any extraneous data near to that character,
the recognition module “normalizes” the image so that it fits a standard
character size. This standardized version of the character is then com-
pared with all of the characters stored in the system's database and a list
is produced of the most probable classification candidates, together with
confidence values indicating how certain the system is of its identifica-
tion in each case. Each character in the database is stored as an image
containing three types of pixels: black, white and grey. Different print
fonts each have their own sets of stored images, in order to prevent the
system from being confused by different fonts.
Handwriting Recognition
Recognising hand-written text is a much more difficult and complex
process than dealing with printed characters. In the early years of charac-
ter recognition there were no mass-market applications for a recognition
technology that could cope with hand-written characters and therefore
there was no serious demand for such systems. But with the advent of
handheld PDAs
3
and other types of portable computers, instantaneous
handwriting recognition has become an important field of research.
The most obvious problem in handwriting recognition lies in the
huge variation in personal writing styles. And many people also exhibit
a lot of variation within their own individual writing style. These vari-
ations can be related to the type of writing equipment (ballpoint pen,
fountain pen, felt-tipped pen, etc.) and to the writer's situation, for ex-
ample whether he is writing on his lap in a moving vehicle or sitting at a
desk in his office. Such differences in style
within
the writings of any one
person encourage the idea that handwriting recognition systems should
be adaptive, so that the system learns the personal writing styles of its
own users.
Handwritten characters vary enormously in shape, due to various
factors:
1. The underlying geometrical properties of the characters vary from
one writer to another, for example the relative positions and sizes
of the strokes, the height-to-width ratios of the characters, and the
general slant of the writing;
3
Personal Digital Assistants.
