Recognition of online handwritten Telugu stroke by detected dominant points using curvature estimation
Srilakshmi Inuganti, Rajeswara Rao Ramisetty · International Journal of Data Analysis Techniques and Strategies · 2022
Online handwritten Telugu character is a mix of strokes, which are from pen-down to pen-up positions. The preliminary objective of feature extractions (FE) is to distinguish the stroke from other strokes. In this paper, we propose a FE method for Telugu strokes by utilising dominant points (DP). This is a non-parametric approach. The procedure initially defines the regions of support (ROS) for each coordinate as per the local properties. With this ROS, the curvature is estimated for every point on the curves and also is utilised to gauge DP. The points encompassing local maximum curvatures are stated as DP. The proposed feature also includes the direction between consecutive DPs of the stroke. The proposed mechanism is verified with HP-Lab data available in the UNIPEN format as it encompasses Telugu characters. It is perceived as of the outcomes that the proposed feature enhances recognition accuracy over the chosen dataset.