|
This
talk describes the principles, methods and practices of printed
Chinese character recognition as described in the reference below.
A stroke structure based approach is use to address the difficult
problems facing Chinese OCR systems. We will focus primarily on
the feature extraction and statistical classifier design modules.
In
particular, the following subjects or components will be touched
on:
Structural analysis and statistical pattern recognition,
Information entropy theory for feature extraction
Types of features extracted
A Minimum Distance Classifier
Tree Classification Design
This
talk is based on:
Chapter 11 of book <> by H. Bunke P.S.P.Wang, World Scientific
1997. A copy of the Chapter is available for copying outside of
room 4405.
|
