Robust Handwriting Estimator from Two Forearm Muscles Activities

摘要

This paper aims to reconstruct handwriting letters only from two forearm ElectroMyoGraphy signals (EMG) activity. Previously, handwriting parametric models have been proposed for this purpose with some success. However these models presents a very high parameters variation due to the problem of EMG variability and its unpredictable propriety. To improve reconstruction accuracy, we developed interval observer, which allows to fuse two Luenberger observers, the first was implemented for upper parameters and the second for lower ones. The proposed method acting as a true handwriting predictor, has shown improvement over the previously proposed techniques, it can be applied with high variability of results across subjects. Moreover, handwriting interval observer model is robust and appropriate for different applications, as: myoelectric prostheses, clinical rehabilitation or even military applications, etc.