Offline Thai hanwritten character recognition system has
been an intensive topic for many researchers. Although many
effiient methods of character recognition are proposed but
many problems still remain to be solved. From the study, we
found tht the previous recognition system [1] cannot support
various styles of writing of Thai handwritten character. To
improve the system, we propose neural network technique
hybrid to th heuristic rules to handle various style of writing
and to fur hmproving the accuracy of our recognition
system emphasize on some problematic feature such as zigzag
feature.