| 摘要 |
A method and system for multistroked oriental-language character-recognition by codifying sequential character strokes, uses a direction code (using two digits and representing four directions) for each stroke's travel direction, and a position code (using two digits and representing four positions) for each stroke's relative position of its starting point relative to the ending point of its previous stroke. Strokes which might have more than one direction at different instants of handwriting are codified by substituting one or two digits in their direction code with one or two replacement digits, for e.g., 'x'. Similarly, strokes which might have more than one relative position at different instants of handwriting are codified by substituting one or two digits in their position code with one or two replacement digits, for e.g., 'x'. A code, for a character, which consists of a sequence of the direction code and the relative position code is called a DXWC. A database which consists of DXWCs can be established by training the computer system using the described method for a big set of characters. A new particular DXWC (called NewC), which may not have any replacement code symbols because of having only a single sample, can be compared with any existing DXWC (called OldC) in the database to achieve a recognition of a new handwritten character. OldCs which might have an 'x' type codes symbols or digits are compared ignoring the 'x' digits.
|
