A New Technique for Segmentation of Handwritten Numerical Strings of Bangla Language

Segmentation of handwritten input into individual characters is a crucial step in connected handwriting recognition systems. In this paper we propose a robust scheme to segment handwritten Bangla numbers (numerical strings) against the variability involved in the writing style of different individuals. The segmentation of digits from a number is usually very tricky, as the digits in a Bangla number are seldom vertically separable. We have introduced the concept of Degenerated Lower Chain (DLC) for this purpose. The DLC method was proved efficient in case of segmenting handwriting digits in our experiments. Ten pages of handwritten Bangla numerical strings containing 2000 individual digits that construct 700 numbers written by five different writers of variable ages were segmented by the developed system. The system achieves more than 90% segmentation accuracy on average.