| As a hotspot of pattern recognition field, the On Line Chinese Character Recognition (OLCCR) has grown into an important computer intelligent interface technology. OLCCR has a wide foreground in application because it accords with human's nature handwriting habit as well as it is real time and convenient in operating. However, there are some difficulties in Chinese Character recognition because of handwritten Chinese Characters'characteristics, such as the huge set, handwriting distortion and imprecise feature extraction and so on. The existing theories and algorithms still have some limitations in solving these difficulties.A novel segment extraction algorithm based on polygonal approximation for OLCCR is presented. With this method, the point with the smallest interior angle is detected and the whole stroke is split into two adjacent curves by this point, which is called as a cut-off point or an inflexion. This method has the advantages of less computational complexity and better approximating effect than other methods.In terms of hierarchical recognition, 36 radicals are selected and defined as head radicals according to the relationship between radicals and Chinese characters. Segment delaying algorithm is adopted to extract these head radicals. For the sake of extracting the besiege radicals of besiege structure characters, of which the head radicals may be extracted incorrectly because of stroke disorder, a new method is proposed to classify besiege structure characters into several classes. Character structure detecting method is carried out in order to extract besiege radicals. The problems of detecting besiege structure and extracting besiege radicals get resolved for the first time.The results of experiments show that segments can be extracted accurately, and the recognition rate of head radical extraction hits 98% using segment delaying algorithm. Besides, with character structure detecting method besiege radicals can be effectively extracted too. |
PDF Full Text Request