Segmentation And Recognition Of Handwriting Numeral Strings With Decimal Point

Recognition of handwritten numeral string has been a research focus in the field of pattern recognition, it has broad application prospects in large-scale data statistics and financial field. The handwritten numeral string segmentation and recognition algorithm of the effect is not ideal enough, and the handwritten numeral string segmentation and recognition research is aimed at pure numbers at present, but handwritten numeral string wiht a decimal point have a more widely applied in various fields. These have a bad impact on the application of handwritten digits recognition technology. This paper proposes a segmentation and recognition method of handwritten numeral string with a decimal pointRcognition of the digit string is based on single digit recognition. Because of the many variants handwritten numbers, it is difficult to extract a group of features that can accurately distinguish the digitals. In this paper, by training the convolutional neural network, the network can learn the the inner features automatically so that it can recognise these digital accurately. In order to to evaluate different segmentation schemes during the touching string segmentation process, an evaluation method of credibility of recognition result is put forward.Considering that handwritten numeral string may contain a decimal point, we extract connected component of numeral string. Through the analysis of the size and the relative position of connected components, we extract the decimal point from the numeral string. We extract the skeleton points from the foreground and background of image, and give different weights to these skeleton points. according to these skeleton points, we may find more than one segmentation curve of numeral string. according to the evaluation of the recognition using the neural network, we pick out the best segmentation curve of the string. At the same time, we get the result of string recognition. Experimental results show that this algorithm has a high recognition accuracy of the decimal digit string recognition rate, for the samples in touching digital data set which is constructed by Oliveira et al, the recognition rate is reach to84.47%, for the recognition of handwritten numeral string contains a decimal point, we get recognition rate of92.6%.