Abstract
This study proposes a new method for character recognition which automatically selects key features in Japanese characters. Our new system adopted genetic algorithms (GA) for feature selection, while experts' experience and knowledge have been used in the previous studies for feature selection, due to lack of effective methods and huge space to search. The algorithm was adapted to ETL-8b database and the effectiveness was evaluated. Score of 91.0(%) was obtained for recognition, which supported that the dimensions of character feature can be effectively reduced by GA.
