In an approach for automatic recognition of a character sequence without writing boxes, the sequence is segmented into characters based on spaces between strokes before recognizing each character. The author proposes another approach consisting of two steps. The first step provides candidates of a character category sequence. Boundaries between characters are also determined in this step. This step compares shape of a handwritten stroke with a template. The next step compares stroke positions with templates. The positions are determined by a boundary provided by the first step. Character templates in a recognition dictionary are stored as a tree, where a mode corresponds to a stroke template and a leaf corresponds to a character category. By using the tree, character recognition problem can be regarded as a path search problem in a tree. A leaf of the tree and the root of the tree are connected to allow search for more than one character.
展开▼