Myoelectric prostheses use surface electromyography (EMG) signals to control motorized artificial limb movement and are clinically available to patients with upper extremity amputation. Recent developments in advanced prosthetic arm systems offer the mechanical means to provide life-like artificial limb control. However, there are limited myoelectric control methods that provide intuitive, simultaneous control of multiple prosthetic degrees of freedom (DOF), which is necessary to replicate the coordinated multi-joint movements produced by intact limbs. Recent studies have demonstrated the potential of using intramuscular EMG to provide simultaneous myoelectric control, but have been limited to conventional one-muscle-to-one-function methods.
In this article, we evaluated the ability of linear regression models to decode patterns of muscle co-activation from intramuscular EMG amplitudes and provide simultaneous myoelectric control of a three-DOF wrist and hand system. We compared the performance of linear regression control to conventional myoelectric control methods using intramuscular EMG, which require users to independently modulate muscle contraction in the residual limb (a challenge for some amputees). In this study, able-bodied subjects evaluated the two systems in a virtual Fitts’ Law task. We measured the speed and efficiency of target acquisition as subjects used each DOF (wrist rotation, wrist flexion/extension, and hand open/close) to position a ring-shaped cursor into the target.
Linear regression control using intramuscular EMG successfully provided simultaneous control of the three-DOF wrist and hand system, providing an alternative to conventional myoelectric control methods. Linear regression control allowed for easier activation of multiple simultaneous DOFs than conventional methods and improved performance metrics in tasks requiring use of all three DOFs. However, subjects using linear regression control also often experienced unintended movements when trying to acquire targets requiring use of only one DOF. These results highlight a tradeoff between providing simultaneous control and the ability to isolate individual DOFs when desired.