A template-based recognition system for on-line handwritten characters

New developments on our real-time recognition engine for isolated handwritten characters are presented. This engine is based on approximate Dynamic Time Warping comparisons with prototypes selected by fast, less accurate classification procedures. The error rate it currently obtains on the standard Pendigits task, 0.60%, is significantly lower than both the error rate of other recently published techniques and the one we ob-tain from the recognition engine included in the Microsoft Vista operating system