There are many types of characters, and each type of character has a variety of fonts. The traditional character recognition methods can not make full use of the features of the character itself. Therefore, only one type of character can be recognized and the recognition effect is not ideal. A kind of character recognition method was presented based on repeated character feature extraction and BP neural network. The proposed method extracted more features of the character itself for a single character image, and the character recognition system by taking advantage of the BP neural network selflearning characteristic was designed. Finally, the recognition process simulation with VC programming was completed. The VC simulation showed that the proposed method could recognize many types of characters, the recognition rate was high and the recognition time was short.