HANDWRITTEN CHINESE CHARACTER RECOGNITION SOLUTION BASED ON CONVOLUTIONAL NEURAL NETWORK

Abstract

Currently, because of its many applications, the handwritten character recognition problem is growing in popularity. The task of recognizing handwritten Chinese characters poses more of a challenge than that of recognizing handwritten numerals and English alphabets for a variety of reasons. To solve this problem, the paper presents a handwritten Chinese character recognition solution based on the famous convolutional neural network architecture Lenet-5. Experimental results on CASIA handwritten datasets demonstrate the ability to accurately predict handwritten Chinese characters for a period of fewer than 0.1 seconds.