A new stroke string matching algorithm for stroke-based on-line character recognition

Each character can be regarded as a string of strokes with each stroke being represented by the type of the stroke. The stroke string is constructed according to the writing order of the strokes. After the type of each stroke has been recognized, a matching algorithm is used to compute the distances between the input character and reference characters. The matching process is carried out in two ways, forward and backward. The authors artificially generated a database for simulating the performance upgrading by using two-way matching instead of one-way matching on stroke strings. The simulation shows that two-way matching works much better than one-way matching when some strokes are misrecognized, omitted, or missequenced.<>