Improving Nepali OCR performance by using hybrid recognition approaches

July 2016

Domains

Natural Language Processing Sentiment Analysis

Authors

Abstract

One of the major reasons for the poor recognition rate of Devanagari OCR is the inadequate handling of dika, vowel modifiers and half forms of consonants, conjuncts, and touching characters during segmentation. We attempted to minimize the segmentation errors by reducing the segmentation tasks. In this work, we propose a hybrid OCR system for printed Nepali text using the Random Forest (RF) Machine Learning technique. It incorporates two different approaches of OCR — the Holistic and the Character level recognition. The system first tries to recognize word as a whole; if it is not confident about the word being recognized, then the character level recognition is performed. Histogram Oriented Gradients (HOG) descriptors are used to define a feature vector of a word or character. The recognition rates of approximately 78.87% and 94.80% are achieved for character level recognition method and the Hybrid method respectively.

Keywords

AI Machine Learning Deep Learning

Cite this Publication!

Nirajan Pant and Bal Krishna Bal. (2016). Improving Nepali OCR Performance by Using Hybrid Recognition Approaches.. IISA 2016, The 7th International Conference on Information, Intelligence, Systems and Applications, 13-15 July, 2016, Porto Carras Grand Resort, Chalkidiki, Greece.

Biblography

@INPROCEEDINGS{7785384,
  author={Pant, Nirajan and Bal, Bal Krishna},
  booktitle={2016 7th International Conference on Information, Intelligence, Systems & Applications (IISA)}, 
  title={Improving Nepali OCR performance by using hybrid recognition approaches}, 
  year={2016},
  volume={},
  number={},
  pages={1-6},
  keywords={Image segmentation;Character recognition;Compounds;Optical character recognition software;Feature extraction;Training;Text recognition;Segmentation;Holistic Approach;HOG Feature Descriptor;Classification;Random Forest (RF)},
  doi={10.1109/IISA.2016.7785384}}