Research On Extracting Technology Of Character Stroke Feature Based On Dynamic Information In The Handwriting Identification System

Nowadays, the automatic handwriting verification system can be divided into on-line handwriting verification and off-line handwriting verification according to the data acquisition ways. The development of off-line handwriting verification has not been very perfect so far. The experts of handwriting verification in China Criminal Police University have found out that the effect of verification can be remarkable improved, during the off-line verification, using dynamic information deduced from static information. To utilizing the dynamic information effectively, the connection between the dynamic information and their corresponding character image is explored to find their match relationship.To solve the problem of how to match dot cloud of Chinese characters to their relative image in the handwriting verification system, this thesis proposes an efficient method to avoid a great deal of complex computation. In the method, the information of the special corner in the dot cloud of character and the character image can be achieved quickly and calculated. According to the computing data, the character image is dealt with by using rotation, scaling and translation. Thus the dot cloud of the character can be matched to the character image. Through testing a great deal of sample datum, this method obtains 97.3% of the average matching ratio. Experimental results have met the need of the handwriting verification system.In this research, a practical and simple method is presented to find the pixel width of character strokes. Through the analysis of the general character stroke extracting algorithm, an algorithm for stroke extracting is developed based on the dynamic information. The algorithm can extract the cross character strokes according to the threshold generated by the stroke pixel's width. Finally, a software system has been developed using the application language VC++. The software is currently testing by China Criminal Police University, and obtains the satisfying result.