EMG/EEG Based Control Scheme of Rehabilitation Robots for Stroke Patients: A Systematic Review

With the increasing global prevalence of stroke, motor rehabilitation for stroke patients has become a prominent area of interdisciplinary research, involving fields such as artificial intelligence, neuroscience, and rehabilitation engineering. In recent years, closed-loop rehabilitation training based on biological signals, primarily electromyographic (EMG) and electroencephalographic (EEG) signals, has gained considerable attention. Despite the inherent complexity of EMG and EEG signals, recent studies have demonstrated the effectiveness of decoding these signals in healthy individuals. However, due to the presence of lesions in the neuromuscular pathways following a stroke, rehabilitation systems that rely solely on EMG or EEG signals still require further improvements in terms of stability and accuracy. This paper provides a systematic overview of the use of EMG and EEG signals for controlling rehabilitation robots in stroke rehabilitation, focusing on the decoding of intention, cognition, gestures, and motor imagery (MI). Finally, the advantages and limitations of EEG and EMG signals are compared, and the feasibility of integrating both signals to control rehabilitation robots is analyzed.