Handwriting Gesture Recognition using Accelerometers

In this paper, we present an accelerometer-based gesture recognition system, and the results from experiments of applying it for classifying 26 handwritten English alphabets. Large number of complex patterns incurs a problem of high user-dependency. To tackle the problem, we made our system as an adaptive learner so that user-specific recognizers can easily be constructed by incremental learning. Our preliminary results on handwriting gestures showed that our adaptive approach incrementally improves the performance of gesture recognizers and exhibited more stable performance than non-adaptive approach.