Handwritten Numeral Recognition Method Based On Genetic-BP Neural Network

Handwritten numeral recognition is an important research project in the fields of imageprocessing and pattern recognition. As a result of the written factors that make number imagesvery arbitrary, such as: stroke thickness, font size, the inclination of handwritten numbers,which will directly affect the final recognition result of digital characters. So recognition ofhandwritten numbers is the most challenging issues of pattern recognition area.The target of this paper is to study the methods of handwritten number recognition. In thispaper it describes the background and development of handwritten numeral recognition. Alsothe preprocessing of the handwritten number recognition is introduced, which discuss thealgorithms of binary, the smooth character segmentation and refinement. The article introducestwo common feature extraction methods statistical feature extraction and structure featureextraction, and then it uses the characters of coarse grid to pick up the number’s charactersunder the environment of matlab. After then, it uses the genetic neural network to train andrecognize. At last, it compares the results of the genetic neural network and the BP neuralnetwork, then find out the advantages and disadvantages.The experimental results show that this method is not only has a great improvement inaccuracy, the convergence rate has also been improved a lot. But this method still has manyplaces need to be improved. In order to find a more perfect and more efficient handwrittennumerals recognition method, we should work more hard in the later study to promote thedevelopment of handwritten numerals recognition.At the end of this paper we summarize the content. And the direction of what we need todo in the future, try to find a perfect digital handwritten numerals recognition method.