Multi-threaded Multilayer Neural Network for Character Recognition

New techniques are evolving day by day to recognize characters in any image, text, cards, etc. Optical character recognition is a well-accepted technique to recognize characters in scene or optically captured image. In this present work, we have evolved new method which employs feed forward back propagation (FFPB) technique interlaced with multi-threaded multilayer perceptron neural network (MTMLNN) to recognize the characters with speedup in operations and better accuracy in character recognition. In present methodology, we have employed structured similarity index measure (SSIM) in MTMLNN to compare image and sigmoid function for neuron activation with stochastic gradient descent method for optimization, to train the neural network to recognize the input character with better accuracy.