Syllable segmentation of Telugu document images

Abstract

This paper describes Syllable Segmentation method for printed text documents in Telugu, a South Indian language. Syllable segmentation is a fundamental process for various applications namely word segmentation, speech synthesis and information retrieval. The segmentation algorithm is motivated by the structure of the script. The segmentation process is complicated due to inherent ambiguities of the natural language. This paper presents a non-dictionary approach for syllable segmentation in the context of Telugu language, using the syllable width and compared with the existing aspect ratio approach. The experimental results show the segmentation accuracy is more than 99%.