Automatic Detection of Movement Inception During Hand Gesture

Hand movement recognition based on ElectroMyographic signals is a crucial subject frequently addressed using feature extraction and classifiers such as neuronal networks. However, there is no model to allow automatic movement inception using superficial ElectroMyographic signals; in the present paper is presented a solution to the above problem. To understand the beginning of movement permits the start of other systems for different applications such as activation of human-machine interface, wearable devices among others, this allows saving energy in terms of computation cost, because only at the right moment the system will activate. The mathematical model presented uses the principle of a) superficial EMG signal model, b) the entropy, and c) the flow entropy. The synergy between all the above elements enables the determination of the produced information rate to track significant variation and causing sharp peaks in the entropy flow. The occurrence of each positive peak in the entropy flow corresponds to a movement begging and the occurrence time of each negative peak corresponds to a movement termination. The model was validated through an experimental procedure using a wearable device, in this case, it is used the MyoArmBand as device for subjects to perform release and grasp gestures. As a result, it is found that the proposed model can automatically detect the beginning of the movements at approximately 245.9ms. This value of time change depends on the device used due to connection systems and the computing capacity. Moreover, the proposed procedure can be implemented using other wearable devices as smart watches, smart bands, smart t-shirts, or others instruments that measure the electrical activity.