Recognition of noisy numbers using neural network

Abstract

In this paper, noisy numbers have been recognized with the help of artificial neural network. Literatures divulge that recently there are many character recognition algorithms for recognition of handwriting, numbers, and alphabets using artificial neural network. These algorithms used multilayer perceptron neural network and large number of input neurons for the recognition. Here, we propose an efficient supervised single-layer perceptron learning with very few input neurons for training. The proposed recognition algorithm is tested on 8 noisy numbers and yields good level of recognition accuracy. Up to 49.28% noise, this algorithm successfully recognizes 87.5% of the given numbers.