An on-line character recognition system for handwritten alphanumeric characters has been proposed in this thesis. Recognizer extracts characteristic features of characters using turning-points based on 4 axes. This system accepts sequences of strokes from a transparent digitizing device with input assumed to be discrete characters within guide marks. We have defined a new character distance metric using the property of turning-points which are defined against X, Y axes and two diagonal axes. Matching distance of a single stroke as well as relationships among strokes are used to obtain the character matching distance.
Experiments have been conducted on several discrete character sets collected by the handwriting recognition group of a foreign research center. Our recognizer's performance has been compared with the recognition rates of the group. Our system has demonstrated that recognition is robust even for varied writing styles.
