Handwriting recognition using acceleration-based motion detection

An alternative to conventional tablet based handwriting recognition is proposed, whereby the movement of the pen is determined from a pair of accelerometers. These accelerometers are mounted on the side of a normal pen and produce a pair of time varying waveforms. This has prompted the modification of some algorithms which have proved successful in the speech area-namely bandpass filtering and hidden Markov models. Early experiments have shown that the scheme has some potential, with a very simple isolated word task attaining over 96% word accuracy for a single writer