Information Technology Reference
In-Depth Information
based delta learning rule, commonly known as back propagation (of errors) rule.
Back Propagation provides a computationally ef
cient method for changing the
weights in a feed forward network, with differentiable activation function units, to
learn a training set of input-output examples. Being a gradient descent method it
minimizes the total squared error of the output computed by the net. The network is
trained by supervised learning method. The aim is to train the network to achieve a
balance between the ability to respond correctly to the input characters that are used
for training and the ability to provide good responses to the input that are similar.
The error of the output computed by network is minimized by a gradient descent
method known as Back Propagation or Generalized Delta Rule.
Outline of this chapter is as follows: Sect. 2 briefs some related work already
done so far by the researchers in this
field. Section 3 presents the motivation behind
this work and various challenges faced during the process of recognition. Section 4
describes the steps involved in the OCR experiment. Section 5 explains various
preprocessing techniques employed to produce good quality image. The feature
extraction technique adopted in this work is explained in Sect. 6 . The process of
Neural Network Training sample preparation is described in Sect. 7 . Section 8
presents the recognition process and the experimental conditions in detail. Section 9
deals with implementation and functional details of the character recognition
experiment. Discussion of results and interpretations are described in Sect. 10 and
finally, the chapter is concluded in Sect. 11 which also presents the future path for
continual work in this
eld.
2 Related Work
A lot of research work had been done and is still being done in character recog-
nition for various languages. OCR is categorized into two classes, for printed
characters and for handwritten characters. Compared to OCR for printed characters,
very limited work can be traced for handwritten character recognition (Desai 2010 ).
Preprocessing is the preliminary step of OCR, which transforms the data into a
format that will be more easily and effectively processed. The main objective of the
preprocessing stage is to normalize and remove variations that would otherwise
complicate the classi
cation and reduce the recognition rate (Alginahi 2010 ). The
use of preprocessing techniques may enhance a document image preparing it for the
next stage in a character recognition system. Thresholding, Noise Removal, Size
Normalization, De-skewing and Slant Correction, Thinning and Skeletonization are
the various pre-processing techniques that have been employed by various
researchers in an attempt to increase the performance of the recognition process.
The Otsu method (Otsu 1979 ) is one of the widely used techniques used to
convert a grey-level image into a binary image then calculates the optimum
threshold separating those two classes so that their combined spread (intra-class
variance) is minimal (Alginahi 2010 ).
Search WWH ::




Custom Search