%0 Journal Article
%T OFF-LINE MIXED DEVNAGRI NUMERALS RECOGNITION USING ARTIFICIAL NEURAL NETWORK
%A Patil S.B. and Sinha G.R.
%J Advances in Computational Research
%D 2012
%I Bioinfo Publications
%X In this paper we are collecting 100 Devnagri numerals from 10 different persons belonging to three different states of India. The data collected in a plane paper is scanned in the form of a bit map image. The collected data has to undergo the preprocessing steps first and then some morphological operations like opening, edge detection, dilation, hole filling and numerals detection are performed. After numerals detection the boundary of the numerals in row, are calculated and stored in the database. Calculating the total number of numerals in a row, the individual width and height of each numeral are measured. The features can be extracted by three steps; first the extreme coordinates of the numerals can be measured, then grabbing the numerals into grids and finally numerals digitization. Digitized numerals are then further trained with multilayer neural network. The proposed research work gives us 100 % accuracy and hence recognized all 100 numerals correctly.
%K Devnagri
%K numerals detection
%K edge detection
%K dilation
%K grids
%K neural network.
%U http://bioinfopublication.org/viewhtml.php?artid=BIA0000284