Off-line Handwritten Chinese Character Recognition Based On Stroke Reconstruction

In the Chinese history of 5,000 years, Chinese characters as the carrier of memory, has been playing an irreplaceable role in the development of Chinese civilization. Nowadays with the rapid development of technology, especially the need of Chinese character informatization, Offline handwritten Chinese character recognition has always been identified as an important research area, Experts associated with it has did a lot of research, Put forward many constructive comments and algorithms., and made a lot of good results.Feature extraction as an extremely important step in the identification process gets a lot of researchers'attention, because feature selection, extraction algorithm are good or bad will directly affect the identification results. But due to off-line Chinese character has the features as :style of writing differently,large arbitrary and structure complicated Etc, and off-line characters are lack of the information about stroke position,stroke order Etc, so extracting the stability structural features is a research goal for the researchers.This paper studied the main points of off-line Chinese Character recognition, focuses on analyzing various existing stroke(horizontal,vertical,left diagonal and right diagonal ) extraction, and proposed a new stroke extraction algorithm base on the summarize previous stroke extraction algorithm. First ,this algorithm set a rectangle with the average of stroke width, using stroke statistical information in this rectangle to determine scanning direction of the window, thus to extract the strokes. Compared with the previous stroke extraction algorithm, the algorithm proposed in this paper is more steady and effective.As the use of the proposed stroke extraction algorithm will contain some redundant stroke information, And in the stroke extraction process may cause the stroke fracture, To solve these problems this paper also present a reconstruction algorithm to fitting,reconstruction the strokes that has been extracted, to achieve the desired stroke extraction result. The experiments have fully proved the effectiveness of the proposed algorithm...