Simultaneous classification of hand and wrist motions using myoelectric interface: Beyond subject specificity

Decoding simultaneous movements in the context of myoelectric control is becoming increasingly popular, because it is a more intuitive and natural way by which humans perform daily life activities. Current decoding techniques require the use of a calibration phase, and also on the use of machine learning algorithms in order to build the decoder model, and hence they are subject-specific. In this paper, we propose a unique subject-independent based decoding model, which is devoid of the calibration procedures required to train the decoder. The idea is to develop a model to decode two degrees of freedom involving the wrist and the hand, and incorporating both individual and combined motions. A set of experiments are performed in order to acquire electromyogram (EMG) signals for the entire set of motions. A hierarchical-decision tree approach is devised to build the model, by analyzing the relative activity patterns of the principal components of muscle activity in both individual and combined motions. The model is tested in a real-Time scenario by means of a virtual graphical environment, and its performance is quantified. The results are promising, and indicate its capability to perform both individual and simultaneous motions.