A Real-Time Line Segmentation Algorithm for An Offline Overlapped Handwritten Jawi Character Recognition Chip

Abstract

Overlapped characters between upper and lower lines can cause a major problem in line segmentation of cursive characters such as Jawi (similar to Arabic characters). In this paper we will discuss our fast approach using a tangent value to find a separation point (SP) for accurate line segmentation without data loss. This approach will be used to design a dedicated Jawi Character Recognition Chip with the objective of obtaining 90% recognition accuracy.