Motor Imagery Analysis of EEG Signals Using a Low-Cost Ankle Exoskeleton

The combined use of a brain-machine interface (BMI) and a robotic device, such as an exoskeleton, is one of the state of the art techniques in neurorehabilitation. Among the control methods employed for BMIs, the paradigm of motor imagery is widely used. To achieve this control it is necessary to define an electroencephalographic (EEG) signal decoding algorithm for the extraction of optimal features. The latter is difficult due to the non-stationary character of the EEG signal and the artefacts that can interfere in it. This study focuses on the analysis and processing of EEG signals from two healthy subjects to develop an algorithm for the decoding of motor imagery (MI). As a method for feature extraction, PSD (Power Spectral Density) will be used. For this purpose, the Welch technique, which is an approach measure of the PSD, is applied. This algorithm has been tested for two electrode combinations. In addition, the EEG signals are conditioned by the movement generated in the foot due to an ankle exoskeleton used. This exoskeleton has been designed with a low-cost philosophy and allows dorsiflexion and plantar flexion movements. The result of the MI decoding algorithm for the first combination was an average accuracy value of 68.8% and for the second combination 75%, providing an important baseline for future research in this area.