Stroke - Database Design for Online Handwriting Recognition in Bangla

Handwriting recognition is a difficult task because of the variability involved in the writing styles of different individuals. This gets more complicated in Asian script for its large character set and presence of modified and compound characters. This paper presents a scheme for the online handwriting recognition of one of the major Indic script, Bangla. Online handwriting recognition refers to the problem of interpretation of handwriting input captured as a stream of pen positions using a digitizer or other pen position sensor. Strokes are extracted from the characters. The sequential and dynamical information obtained from the pen movements on the writing pads are used as features in the proposed scheme. These features computed from the strokes, and fed to the MLP classifier for recognition. The characters are identified from its constituent strokes. The system was tested on 21372 Bangla character data and obtained 92.92% accuracy with top 3 choices.