For access to this article, please select a purchase option:
IET members benefit from discounts to all IET publications and free access to E&T Magazine. If you are an IET member, log in to your account and the discounts will automatically be applied.
Your recommendation has been sent to your librarian.
Artificial neural networks (ANNs) have been recognized as universal estimators and widely used in a range of fields starting from computer science, medical, neuroscience, engineering, remote sensing, artificial intelligence to business and stock markets. In this study the effectiveness of ANNs for recognition of character pattern as an example has been studied. A programming code has been developed to build ANN model system that utilized feedforward backpropagation methodology in learning session and subsequently recognized several predefined alphabet characters. Based on computational resources A 7×5 matrix bit map was achieved for a number of characters such as A, B, C, D, E and F. Backpropagation training was used to adjust the weights of the branches connecting the neuron layers. Thirty-five digitized values (0, 1) against each character were fed to the model as input variables and fetched to thirty-five input neurons of the as-modeled ANN system. The output was considered as specified codes for each character such as 01010, 01010, 01100, 01101, 01110 and 01111 respectively. The training continued till the predefined tolerance limit reached to less than 0.0001. Forward run of the ANN processing was observed capable enough to recognize some irregular pattern in the character. The results showed that with some modifications in the bit pattern the ANN model system could recognize the pattern of the exact character with maximum 37.5 % deformation. Such a model can be exploited to further advancement in pattern recognition system developments.
Inspec keywords: artificial intelligence; backpropagation; feedforward neural nets; image recognition; pattern recognition
Subjects: Computer vision and image processing techniques; Neural nets; Image recognition; Knowledge based systems