Representation, Extraction of Nodal Features of DevNagri Letters.
Rajiv Kapoor, Deepak Bagai, Tara Singh Kamal · Indian Conference on Computer Vision, Graphics and Image Processing · 2002
This paper proposes a method for the recognition of handwritten DevNagri characters. Recognition has been done using an HMM-based approach, using junction points of a character as the main feature. The character has been divided into 3 major Zones (Upper, Middle, and Lower). This paper explains the method for middle zone which depends upon three major features i.e. number of paths, directions of paths, & region of the node. Tolerance of feature points has been proposed as an additional important parameter. Heuristic features taken up from the existing methods have been redefined and refined for this application. The connectivity features which give meaning to the whole word have been redefined, excluding number of Matra endpoints, Matra type (curve/line). Finally every character has been recognized using HMM. The proposed method has been tested for the sample-words taken from 100 persons.