02-09-2017, 10:44 AM
an artificial character recognition system based on artificial neural networks (RNAs). The ANN is trained using the Backward Propagation algorithm. In the proposed system, each letter written in English is represented by binary numbers that are used as input for a simple feature extraction system whose output, in addition to the input, is fed to an ANN. Later, the Feed Forward Algorithm gives an idea of the operation of a neural network followed by the Rear Propulsion Algorithm that compromises the Training, the Calculation of Errors and the Modification of Weights.
The recognition of optical characters is known to be one of the first applications of Artificial Neural Networks, which partially emulate human thinking in the domain of artificial intelligence. The paper is expected to serve as a resource for students and amateur researchers in recognizing patterns, neural networks and related disciplines. It has been 50 years since the idea arose that computing systems can be made in the replication of biological neural networks. However, the development of this branch of science made the improvement of these systems only possible in the last 25-30 years. Today, neural computing is a very extensive and separate science. Its solid theory base made possible its use to solve many types of problems in artificial computing, thus improving the experience of science. Neural networks are commonly used to solve sample recognition problems. One of them is character recognition. The solution to this problem is one of the easiest implementations of neural networks. With the help of Matlab's neural network toolkit, we try to recognize printed characters and manuscripts by projecting them into networks of different size (5 × 7, 7 × 11, 9 × 13). The results showed that the accuracy of the character recognition depends on the resolution of the character projection. In addition, we realized that not all writing styles can be recognized using the same network with the same precision. This demonstrates that the variety of human handwriting habits can not be completely covered with a neural network.
The recognition of optical characters is known to be one of the first applications of Artificial Neural Networks, which partially emulate human thinking in the domain of artificial intelligence. The paper is expected to serve as a resource for students and amateur researchers in recognizing patterns, neural networks and related disciplines. It has been 50 years since the idea arose that computing systems can be made in the replication of biological neural networks. However, the development of this branch of science made the improvement of these systems only possible in the last 25-30 years. Today, neural computing is a very extensive and separate science. Its solid theory base made possible its use to solve many types of problems in artificial computing, thus improving the experience of science. Neural networks are commonly used to solve sample recognition problems. One of them is character recognition. The solution to this problem is one of the easiest implementations of neural networks. With the help of Matlab's neural network toolkit, we try to recognize printed characters and manuscripts by projecting them into networks of different size (5 × 7, 7 × 11, 9 × 13). The results showed that the accuracy of the character recognition depends on the resolution of the character projection. In addition, we realized that not all writing styles can be recognized using the same network with the same precision. This demonstrates that the variety of human handwriting habits can not be completely covered with a neural network.