In this paper, a new character segmentation algorithm for dealing with off-line handwritten script recognition is presented. The X-axis projection, Y-axis projection and geometric classes techniques used by the algorithm proves to be successful in segmenting normal handwriting with a success rate of 93.5%. As a result of this development, detailed understanding of geometric classes of English characters and the difficult cases in segmentation was gained. Although the algorithm works quite well with a randomly chosen sample, results of a detailed analysis may shed new light into the tuning of the algorithm especially for segmenting the identified difficult cases.
展开▼
