ADAPTIVE ON-LINE HANDWRITING RECOGNITION.

Abstract

For the automatic recognition of handwriting entered via a tablet terminal, the range of pattern recognition problems can be described by the severity of letter segmentation required. Discrete characters written in boxes require no letter segmentation. Spaced discrete characters require spatial letter segmentation. The more difficult problems require the interaction of letter segmentation and recognition. These problems include overlapping discretely written characters, pure cursive writing, and mixed cursive and discrete writing. The use of dynamic programming to perform elastic (nonlinear) matching of two time sequences of measurements, found useful in automatic speech recognition, is applied successfully to handwriting recognition.