Dynamical encoding of cursive handwriting

Online cursive handwriting is considered as a slow modulation of an underlying cycloidal motion, Two dimensional oscillation, with a constant linear drift, describe the general pen motion. The dynamical equations describing the oscillations are coupled through fixed ratios of the angular velocities and phase lags. The entire process is thus viewed as an almost constant vertical oscillatory movement with changing horizontal velocity phase lag. An estimation scheme of the cycloidal motion parameters is presented. In the estimation process the instantaneous amplitude and phase lag of the horizontal velocity are calculated and quantized. The result is a many to one mapping from the continuous pen movements to discrete motor control symbols. Using this motor control representation word spotting and matching is performed successfully. These results suggest the potential of our dynamic representation for complete cursive handwriting recognition.