On-line handwriting recognition using physics-based shape metamorphosis

We present a new technique based on shape metamorphosis for on-line recognition of handwritten words and simple shapes in a user-dependent setting. The approach includes a segmentation method that does not try to locate letters but instead performs the significantly easier task of locating corners and some key low curvature points. This is part of the method's strategy to see the word as a generic on-line shape. The segmentation points are used to model a cursive word or a hand-drawn line figure by pieces of wire. Shape metamorphosis occurs through stretching and bending of the artificial wire. The amount of energy spent in morphing one shape to another is used as a dissimilarity measure. For any two given shapes an optimal morph can be computed in quadratic time by constraining the metamorphosis to the segmentation points of these shapes. Experiments with multiple subjects indicate that the method can handle collectively cursive words and hand-drawn line figures, both useful forms of communication in pen-based computing.