On-line Handwritten Lao Character Recognition by using Dynamic Programming Matching

This paper describes a method for online handwritten Lao character recognition by using dynamic programming matching (DPM). It extracts feature points along the way from pen-down to pen-up, then uses DPM to match those feature points with the feature points for a template pattern of each character class and obtains a similarity for each character class. It selects the character class with the largest similarity as the recognition result. We also compare the recognition method by using DPM with that by the linear-time elastic matching (LTM). An evaluation on the Lao character pattern database shows the result that the speed gain by LTM is small and DPM brings a better and more accuracy recognition rate.