The Research On The Recognition Of Bangladesh Handwritten Numerals

Devijver and Kittler defined feature extraction as: "relevant information, which extracted from the original data used for classification, make new expression patterns reduce difference of class, at the same time increase the differences between the class". It is a very important part of handwritten digital identification system and has a direct impact on the recognition rate system. Neural networks are widely used in pattern recognition because of stronger self-organization learning, fault-tolerant, robust and nonlinear processing capacity. This paper apply research on Bangla handwritten numerals on the basis of structural features and neural network. Experimental results show that the method is of high reliability in recognizing Bangla handwritten numerals and meets the practical requirements and has some theoretical and practical value.The main content and structure of this dissertation are as follows:Image preprocessing on Bangla handwritten numeral images is implemented, including image smoothing, binarization, stroke width normalization, character size normalization, etc. The dissertation propose a new recognition method for Handwriting Character with Directional features based on character statistical features. The method has good Anti-noise performance but has certain limitations. The method of recognizing rough grid features based on neural network have Strong anti-noise ability. This paper using a mixture of the two methods. According to the special structure of Bangla handwritten numerals, open-loop and closed-loop features are employed. These features are able to describe each Bangla numeral category and distinguish one from others, which reduce the dimensions effectively and greatly improve the efficiency of the classifier. Multiple classifier systems is used as classification system of Bangla handwritten numerals recognition, and the dissertation describe the corresponding algorithms and programming steps. Experiments show that the recognition system reached higher recognition rate.