Efficient Low-resolution Character Recognition Using Sub-machine-code Genetic Programming
Giovanni Adorni, Stefano Cagnoni, Marco Gori, Monica Mordonini · 2003
The paper describes an approach to low-resolution character recognition for real-time applications based on a set of binary classifiers designed by means of Sub-machine-code Genetic Programming (SmcGP). SmcGP is a type of GP that interprets long integers as bit strings to achieve SIMD processing on traditional sequential computers. The method was tested on an extensive set of very low-resolution binary patterns (of size 13 × 8 pixels) that represent digits from 0 to 9. Ten binary classifiers were designed, each corresponding to a pattern class. In case of no response by any of the classifiers, a reference LVQ classifier was used. The paper compares the resulting classifier with a reference classifier, showing an almost 10-fold improvement in speed, at the price of a slightly lower accuracy. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.