Improvement on the Digit Recognition Technique of Backpropagation Neural Networks

Abstract

Digit recognition based on backpropagation neural networks, as an important application of pattern recognition, attracted much attention. Although it has the advantages of parallel calculation, high error-tolerance, and learning capability, better recognition effects can only be achieved with some specific fixed format inputs of the digit image. Moreover, the recognition rate can not be improved obviously even if digits were trained with various format samples. This is because the accuracy of recognition is also directly affected by the digit image preprocessing ability. Here using Matlab software, the digit image was enhanced by resizing and neutral-rotating the extracted digit image before recognition, which improved the digit recognition capability of the backpropagation neural network for practical applications. This method may also be helpful for recognition of other complex patterns with backpropagation neural networks.