Copper Plate Image Character Recognition System Using Extreme Deep Learning Machine

  • Indra Gandhi R et al.

Abstract

Tamil is conceivably the oldest language on the earth, spoken in Tamil Nadu, South India, which is
assimilated from Brahmi Script. The primary wellspring of data about history are the stone
engravings. OCR helps in digitizing Tamil contents from the old era to the most recent, making its
access simply through Internet. Antiquated Tamil character acknowledgment from stone engraving is
a test because of the huge inconsistencies of composing style. Efficient feature extraction and
selection is basic steps to understand the Ancient Tamil character recognition framework. The
principle objective for an OCR is to build the recognition rate and to adapt to the low quality of
scanned images. Investigating the best feature extraction methods and choosing a proper selection
algorithm and classification techniqueslead to prevalent recognition precision and low computational
overhead. This paper presents, first time, a new methodology by adopting two approaches. The first is
Extreme Deep Learning Machine (EDLM), algorithm for classification that has a short processing
time. Also, EDLM keeps away from numerous troubles looked by gradient based learning strategies,
such as learning epochs and local minima. The subsequent calculation is a Statistical based feature
selection that has better convergence and spread of measures. EDLM accentuates non-dominated
elucidations and explicit diversity preservation mechanism.Comparison with the experimental results
of other methodologies revealed the proficiency of the proposed system and demonstrated that the
feature selection approach increased the accuracy of the classification process.

Published
2020-02-16
Section
Articles