I. Methasate and S. Sae-Tang (Thailand)
On-line Character Recognition, Thai, HMM, Stroke segmentation.
This paper describes the On-line Thai character
recognition system which used the segmented stroke
information by Hidden Markov Model (HMM). Firstly,
the stroke sequences are smoothened and unnecessary
coordinates are removed. Secondly, the strokes are
divided into sub-stroke and then merge some short stroke
if necessary. In this step, the features of sub-stroke are
extracted. Finally, the feature vectors are recognized by
Hidden Markov Model. In applying free-hand Thai single
characters to the system, it gives the average recognition
rates of 92.17 %.