24-point game based on BP neural network : A smart game designed for children

Abstract

This paper introduces a 24-point game based on BP (Back Propagation) neural network. This game enables two children to play together and have a contest, which will help the children to strengthen their number recognizing and mathematics ability. The training set consists of 2340 28*28 pixels black-and-white images. Hough transform and morphology are used to preprocess, then we segment the image and extract the feature. We construct a BP neural network by MATLAB neural network toolbox. Input the image processed by the front end into the network to recognize the characters of the poker card. After recognizing the 4 numbers, all the right answers to the 24-point game will be calculated and shown to the players by the system. At the same time, the system can count the time the players spend/cost and give a feedback.