A Neural Approach to Cursive Handwritten Character Recognition Using Features Extracted from Binarization Technique - Complex System Modelling and Control Through Intelligent Soft Computations - page 743

Fig. 11 a Binary image of character ' c ' b resized binary image of character ' c ' c binary matrix

representation, and d reshaped binary matrix or feature vector of character ' c '

because there might be only one or two values, which are signi cant to recognize a

particular character. The performance of the recognition system relies much on the
quality of the features extracted as well as on the selected classi er itself.

12
matrixes as shown in Fig. 11 b. Each cropped and size normalized character image in
binary format is traced vertically column wise. A white pixel is represented by

The binary image of character ' c ' as shown in Fig. 11 a is resized to 15 ×

' 0 '

and a black pixel is represented by ' 1 ' and the binary matrix representation of

' ' ×

character c is shown in Fig. 11 c. This binary matrix of size 15 12 is then reshaped

× ' '

in a row first manner to a binary matrix of size 180 1 by using reshape method of

MATLAB and is shown in Fig. 11 d. The column vector of size 180 × 1 as shown in

Fig. 11 d is a feature vector for the character image ' c ' shown in Fig. 11 a.

7 Sample Preparation for Neural Network Training

The feature vector of a single character is a column vector of size 180 × 1. One such

feature vector of character ' c ' is shown in Fig. 11 d. Similarly, the feature vectors of

all the 26 characters (a z) are created in the form of binary column matrix of size

-

180 1 each. All these 26 feature vectors are combined to form a sample which is a

binary matrix of size 180

×

× 26 as shown in Fig. 12 .

26, there are 26 columns representing 26 characters
of the English Language and each column represents feature vector of length
180

In this matrix of size 180 ×

× 1 of a single character e.g. first column represents feature vector of character

' a ' , second column represents feature vector of character ' b ' , third column repre-

sents feature vector of character and so on.

For sample creation, 1,300 characters were gathered where each writer con-
tributed 5 samples of the complete English alphabet (a

' c '

z). After pre-processing,
these samples were considered for training such that each sample was consisting of
26 characters (a

-

- z).