The feature extraction module contains structural feature
extraction sub module and statistical feature extraction sub
module. For training and testing purpose, we collected a large set
of English handwritten characters from the different peoples of
different age groups. The total 25000 character images is divided
into two groups. Total 18000 of samples were treated as training
samples and other 7000 ofsamples were treated as test samples.
The character images from the original database are rearranged in
the test sets and in the learning sets. All samples were normalized
to 64 X 64 bitmap, which was further divided into 16 X 16 equal
sized blocks. From these blocks, we obtained 128 X 128 grey
level pixels presented with real numbers in [-1,1], intervals. The
recognition results for the proposed method are relatively high.
The recognition results are 99.9% for English handwritten
characters.