Pre-processing techniques for Tamil online handwritten character recognition

In the modern era, due to the rise of the information technology field, human-computer interfacing devices such as a tablet, pc, and smartphones are becoming more popular [1]. This technological advancement led to the advancement of online handwriting character identification system-based applications. Handwritten characters are recorded in digital form by writing on an electronic surface with a digital pen/stylus. Online handwriting recognition systems face several challenges, out of which some could be tackled in the pre-processing part and few can be resolved in the post-processing part. In this paper, the pre-processing steps for the Tamil handwritten character recognition have been addressed. Due to the high cursiveness property and similar shape of some characters in the Tamil script, it is necessary to work on these challenges at the pre-processing level. Therefore, the pre-processing techniques discussed in this study are; duplicate point removal, size-normalization, missing point interpolation, and points resampling. A total of 10 Tamil consonants samples from the hpl-Tamil-iso-char dataset have been utilized which was collected by HP Labs.