Printed New Tai Lue Character Recognition Technology Study

At present, the informatization of the language and character of YunNan XiShuangBanNa Dai nationality has lagged behind. The development of the New Tai Lue recognition technology is important to develop the imformatization of the language and character of YunNan XiShuangBanNa Dai nationality and study the history, culture and preserve the text information and ancient literature of Dai. This article d introduces two types of recognition method of printed new Tai Lue, they are BP neural network and convolution neural network. These two ways are similar in process, the difference lies in the classifier design, This article introduces two ways of processing this two parts, and emphasis and difficulty of it. Solving process is as follows:In the image pre-processing stage, Firstly, we choose scanner as the best tool of image input. Secondly, we come up with series algorithm of Tilt correction, binarization and thinning operation. Lastly, we propose possible new Tai Lue character segmentation method, starting from Formation and character segmentation difficulty of new Dai.In the recognition stage, mainly introduces two types of recognition method based on the BP neural network and convolution neural network. Among them, we select local gray feature as vector features when building BP neural network, and use double hidden layer to construct network. The final recognition rate is 98.03%. Convolution neural network using the network structure LeNet5, the final recognition rate is 98.302%.The last of the article analysis advantages and disadvantages of two types of recognition, and difficulties and problems need to be solved in the process of recognition. Finally, the article comes up with the improvement program and the future direction of New Tai Lue research...