Feature Recognition Method for Number Recognition and Number Recognition Circuit Using Neural Network

摘要

The present invention discloses a number recognition circuit using a feature amount calculation method and a neural network for number recognition. The method comprises dividing a numerical image consisting of N × M into n × m sub-images non-linearly so that the number of black pixels is uniformly distributed in the vertical direction and the horizontal direction, and black pixelated white pixels in the sub-image. Calculate the value of the horizontal direction plus the length in the vertical direction, and the horizontal length and the vertical length with respect to the black and white pixels, respectively. And a feature amount calculation step of calculating the feature amount. The circuit inputs the feature amount calculated by the specific amount calculation method and outputs the recognized result. Therefore, the present invention can increase the recognition rate by accurately calculating the feature amount of the numeric image.