Glyph-based recognition of offline handwritten Telugu characters: GBRoOHTC

摘要

Recognizing offline handwritten Telugu characters from digitized document images is very challenging. In this paper, we propose a novel approach of hybrid feature extraction and hierarchical classification to recognize the glyphs of offline handwritten Telugu characters. In the proposed method, hybrid features are extracted from the glyphs and the glyphs are recognized using a hierarchical classification system. The hybrid features are extracted on the basis of glyph dimensions, positions of small glyphs, and zone-wise features for glyph classification. We implemented a two-level hierarchical classification method for classifying the glyphs of offline handwritten Telugu characters. The proposed method efficiently recognizes the glyphs of offline handwritten Telugu characters. The overall recognition rate is 88.15%.