Alphanumeric Character Recognition Based on BP Neural Network Classification and Combined Features

This paper puts forward a new method of alphanumeric character recognition based on BP neural network classification and combined features. This method firstly establishes three BP networks respectively for three categories of characters which are classified according to their Euler numbers, with the combination of grid feature and projection feature as the input of each BP network. When recognizing a character, its combined features are fed into the three BP networks simultaneously without the necessity for judging its Euler number. The final recognition result is elaborated by synthetically analyzing the outputs of three BP networks. Experimental results show that the proposed method can effectively improve the recognition ability and efficiency, and has a good property of fault tolerance and robustness. Furthermore, the weight coefficients of combined features for each BP network are optimized, which can further improve the recognition rate.