Handwritten Digit Recognition Based On GANN Algorithm

Handwritten digit recognition (HDR) has a prosperous future in the field of simulating artificial intelligence and computer words processing. Researchers worldwide worked on pattern recognition and proposed many algorithms of image pre-processing and pattern recognition. But none can compare with the recognition ability of human beings. This inspires researchers to improve the image pre-processing algorithm, feature extraction algorithm and recognition algorithm.Based on many references here, a new HDR algorithm of Genetic Artificial Neural Network (GANN) algorithm is proposed. GANN Algorithm is a perfect combination of genetic (GA) and artificial neural network (ANN).It appropriately select one algorithm from GA and ANN to apply according to the output error in the network training. GANN algorithm develops the advantages of global optimization capabilities of GA and local optimization of ANN. It calculates the final result only through the forward network instead of a more complex algorithm.The GANN algorithm extracts a series of structural features of the digits. They are the number of crossing point in vertical centerline, crossing point in vertical one third area, crossing point in vertical two thirds area, crossing point in horizontal one fifth area, crossing point in horizontal four fifths area, inflection-point to left, inflection-point to right, upper extreme point, lower extreme point. These features are used as the input value of GANN.Through analyzing of the training results and recognition results a satisfactory result is obtained. The combination of GA and ANN brings a new idea to researchers. If a further researching on the feature extraction is done and a better algorithm of digit feature is found, a better network recognition effect will be realized.