Complicacy in Electrode Position Shift and Its Solution in sEMG Pattern Recognition: A Review

Upper limb amputation can severely restrict an individual’s ability to perform daily tasks. Myoelectric prostheses aim to restore the functionality of amputated limbs by utilizing signals from the remaining muscles in the residual limb. However, achieving high pattern recognition (PR) accuracy from acquired surface electromyography (sEMG) signals is a complex challenge. Several significant factors—including electrode position shifts during donning and doffing, interuser variability, and issues related to muscle sweating—pose obstacles to attaining an accuracy rate exceeding 90%. These challenges can result in the production of prosthetic hands that do not function as intended. To address these issues, modified machine learning schemes and deep learning algorithms can be implemented to find optimal solutions. This review serves as a valuable resource for researchers, helping them understand the challenges associated with electrode position shifts and the solutions developed to enhance the performance of sEMG-based PR systems. It also reviews the causes of electrode position shifts, techniques for identifying shifts, and methods to mitigate the problems caused by electrode displacement in traditional sEMG-based PR. In addition, the pros and cons of each technique are discussed based on window size, number of electrodes, features, gesture numbers, and types of algorithm performance of each technique. This work will also help researchers identify current research challenges related to electrode position shift problems and the solutions that have been proposed to overcome performance limitations.