Pattern Recognition using Neural Networks

Abstract

Character recognition is just one part of the pattern recognition existing in the world. The major advantage of doing the pattern recognition using artificial intelligence and unsupervised learning is that with the use of correct data set you can teach it to recognize every pattern that exists. Our mind can do pattern recognition or handwritten character deciphering very efficiently and easily, simply because our mind is made up of a large set of neural networks. Doing it in real time is another challenge but we have considered challenges as stages of success rather than considering them as hurdles in achieving our goal. Thus using non biological neural networks we have implemented character recognition using raspberry PI. The problem still exists because of the competition between the efficiency and speed. A camera is used for image acquisition in the real time and after filtering the noisy data and preprocessing it is passed on. Feature extraction is another important part of the procedure in which differentiation between the characters is done. It helps recognizing the characters. The characters in the photo are stored as an image of matrix of pixels. Back propagation algorithm is used to optimize the results achieved so that our output is closer to the results expected. Weights and biases of the neural network are automatically adjusted according to the back propagation which further helps in improving the results