Extraction of features of kinesthetic activities in the upper limb from EEG recordings based on sub-band analysis with wavelet transform for the control of robotic assistance systems

In the literature, findings of movement and force control of non-intrusive robotic assistance systems have been reported, based on the processing of electroencephalogram (EEG) recordings and the conditioning of the control strategy. However, the collection of signals in the cerebral cortex may not represent the set point defined in the cerebellum, particularly in post-cerebral stroke patients. The present study reports a new proposal inspired by human intention of a gross motor action assisted by a robotic platform. For this, a BCI (Brain-Computer Interface) system is used, based on the instrumentation of EEG signals, specifically from the cerebral cortex, and its digital processing using wavelet multiresolution analysis for the detection of features associated with real and imagined kinesthetic tasks, such as: upper limb movement and manipulation of objects. The main result is associated with the position control of a direct current motor and the motion control of a 2 DOF manipulator robot, with a high performance adaptive wavenet PID control for stabilization of non-linear MIMO (Multiple Inputs Multiple Outputs) systems.