Computer recognition of handwritten Japanese characters by wisdom algorithms
A practical system has been developed that can recognise hand-written Japanese characters. The recognition method was achieved by using fluency functions to observe written patterns an d relaxation methods for matching. It has been found that this system has an accuracy of 99.7% and a speed of 10 characters per second on a host machine capable of 15.8 million instructions per second with a parallel subunit as determined by trials on a portion standard data set ‘ETL-8’ for handwritten Japanese characters.