Myoelectric prosthetics

Myoelectric prosthetics are artificial limbs driven by motors and controlled by the electrical activity of the wearer's own remaining muscles. Surface electrodes in the socket pick up electromyographic signals a few hundred microvolts in amplitude, the prosthesis converts their envelope into motor commands, and the user learns to produce the required contractions. The approach dominates powered upper-limb prosthetics, and its persistent problem is not the hand — modern hands are mechanically excellent — but the narrow, noisy control channel available to drive it.

§How it works

Muscle fibres depolarize when they contract, and the summed extracellular field of many motor units is detectable through the skin. A pair of differential electrodes pressed against a residual muscle records this signal; the prosthesis rectifies and smooths it into an amplitude envelope and maps that envelope onto a motor's velocity. Squeeze harder and the hand closes faster.

The classical arrangement is two-site proportional control: one electrode over a flexor group, one over an extensor group, driving a single degree of freedom in opposite directions. A wrist rotator or elbow requires switching modes, usually by co-contracting both muscles or by holding a contraction past a timeout. Each additional joint therefore costs the user a deliberate, non-intuitive switching action, which is why devices with five actuated fingers are commonly operated as though they had one.

Signal quality is the limiting variable. Electrode contact shifts as the socket moves, sweat changes skin impedance, muscle fatigue changes the envelope, and limb position alters which muscles are recruited for the same intended action. A control scheme trained sitting still on a bench often degrades when the arm is raised overhead.

§Control strategies

Direct control maps one muscle site to one function. It is robust, transparent to the user, and severely limited in the number of functions it can address.

Pattern recognition records from an array of electrodes around the residual limb and classifies the spatial pattern of activity to infer which movement the user is attempting, selecting a grip or joint automatically rather than requiring a switching gesture. Commercial pattern-recognition controllers have been available since the 2010s. Laboratory accuracy is high; performance in daily life is degraded by the same electrode-shift and limb-position effects, and users must retrain the classifier periodically.

Regression and continuous decoding treat the problem as estimating joint velocities directly rather than selecting from a menu of classes, allowing simultaneous multi-joint control. This is the same shift from classification to continuous estimation that improved Neural decoding in cortical systems, and it faces the same non-stationarity problem.

Cortical control is sometimes proposed as the alternative. For most amputees it is the wrong tool: the peripheral signal is intact and far easier to access, and a Brain–computer interface adds neurosurgical risk to solve a problem that surgery on the arm can address more cheaply.

§Surgical amplification of the signal

The most consequential advances have come from operating on the residual limb rather than improving the electronics.

Targeted muscle reinnervation (TMR), developed by Todd Kuiken's group and first performed in 2002 on a patient with bilateral shoulder disarticulation, transfers the severed nerves that once controlled the arm onto spare muscle in the chest or upper arm. The reinnervated muscle then contracts when the user thinks about the original movement, converting a nerve signal that had nowhere to go into a recordable, intuitive EMG site.¹ TMR also reduces neuroma and phantom limb pain, and a randomised trial supported its use for pain even in patients not seeking myoelectric control.²

Regenerative peripheral nerve interfaces wrap a transected nerve in a free muscle graft, which the nerve reinnervates and which then acts as a biological amplifier. Implanted electrodes on such grafts have supported real-time individual finger control in upper-limb amputees.³

The agonist-antagonist myoneural interface (AMI), developed in Hugh Herr's group, surgically reconnects opposing muscle pairs in the residual limb so that contraction of one stretches the other, restoring the proprioceptive signalling that ordinary amputation destroys. In a trial of below-knee amputees with AMI surgery, neural control of a powered ankle produced walking speeds and gait biomechanics closer to unimpaired norms than in amputees with standard surgery.⁴

Skeletal attachment through Osseointegration removes the socket entirely and allows electrode leads to pass through the implant rather than across the skin, yielding stable recordings and a path for stimulation back into the nerve. Self-contained arm prostheses using this route have been used at home for years by a small number of patients.⁵

§Development history

1. 1948–1960s
   First myoelectric arms
   Reinhold Reiter demonstrates myoelectric control in Munich; Soviet and Viennese groups field the first clinical devices, and the 'Russian hand' is exhibited internationally.
2. 1960s–1980s
   Clinical adoption
   Two-site proportional control becomes standard; thalidomide-related limb difference in Europe drives paediatric fitting programmes.
3. 2002–2009
   Targeted muscle reinnervation
   Nerve transfers create new intuitive control sites; a case series demonstrates real-time control of multifunction arms.
4. 2007–2015
   Multi-articulating hands
   Independently actuated fingers reach the market, offering many grip patterns but no corresponding increase in control channels.
5. 2013–2020
   Pattern recognition and implanted electrodes
   Commercial pattern-recognition controllers are released; osseointegrated systems with implanted electrodes and nerve stimulation are used long-term outside the laboratory.
6. 2024
   Restored proprioception in gait
   Amputees with surgically reconstructed agonist-antagonist muscle pairs achieve near-normal walking biomechanics under continuous neural control of a powered ankle.

§Abandonment and the case for the hook

Surveys of upper-limb prosthesis users consistently find that a substantial minority stop using their device, with rejection of externally powered arms reported at roughly one in five among adults and higher among children.⁶ The reasons are unglamorous and repeat across studies: weight, lack of sensory feedback, insufficient durability, discomfort, slow donning, poor reliability in wet or dusty conditions, and cost of repair.

Body-powered prostheses — a harness across the shoulders operating a split hook through a cable — are lighter, cheaper by an order of magnitude, essentially unbreakable, and give the user crude force feedback through cable tension, a form of proprioception no commercial myoelectric hand provides. Many experienced amputees prefer them for work and keep a myoelectric hand for social settings. Any honest account of the field has to record that the most advanced device is not always the one people choose.

The mismatch

A multi-articulating hand offers a dozen or more programmed grips. Standard two-site control supplies one degree of freedom at a time. Most of the mechanism's capability is inaccessible in ordinary use, which is why control research has produced more functional gain than hand design.

Cost compounds the problem. Advanced hands run to tens of thousands of dollars before the socket, fitting, and therapy that determine whether the device works, and reimbursement varies sharply between health systems and between civilian and veteran populations within the same country. Most of the world's amputees, concentrated in low- and middle-income countries and in populations affected by conflict and diabetes, have no realistic access to any powered device — the pattern described under Access and inequality. Low-cost 3D-printed powered hands have narrowed the gap for some paediatric users without closing it.

§Limitations

The channel is the constraint. Surface EMG offers a handful of noisy, correlated signals from muscles that were not designed to be independently controlled, and no amount of hand engineering adds control dimensions. Sensory feedback remains the largest functional gap: without it, users watch their hand continuously and cannot modulate grip force by feel, which makes handling fragile objects slow and error-prone. Implanted nerve stimulation restores graded touch in research settings, but no commercial upper-limb prosthesis provides it as of 2026. The contrast with the Cochlear implant, which has delivered an artificial sensory code to hundreds of thousands of people for decades, is instructive: hearing tolerates a crude code, whereas touch used for motor control needs timing and force information the current interfaces cannot supply.

Prosthetic limbs also do not restore the limb. They restore a subset of its functions to a person who must actively operate them, which is why users describe fatigue as a reason for abandonment and why the field's ethical literature resists framing amputation as a solved problem — a point developed under Disability rights and enhancement. A minority of users take the opposite view and treat the prosthesis as a platform to be customized rather than a substitute for a lost hand, the position argued under Morphological freedom, and interchangeable tool attachments and non-anthropomorphic end effectors follow naturally from it.

§Outlook

The near-term direction is surgical and biological rather than mechanical: nerve transfers, regenerative interfaces, and reconstructed muscle pairs that create better signals for existing electronics, combined with implanted rather than surface recording. Bidirectional systems that combine implanted stimulation for touch with skeletal attachment are the plausible next clinical product, following the pattern of Neuroprosthetics generally, where writing information in lags reading it out. Adding channels the body does not already have — vibration, temperature, or grip force mapped onto an unrelated nerve — shades into Sensory augmentation, and users of touch-restoring systems have reported that the mapping need not be anatomically faithful to be usable.

Two further prospects sit outside the incremental path. Powered Powered exoskeletons devices and prosthetic limbs are converging in actuator technology, and competitive prosthetic athletics has already forced the question of when a replacement becomes an advantage, examined under Enhancement in sport and Human enhancement. And biological replacement through Limb regeneration or engineered tissue would make the entire control problem moot; nothing in the mammalian regeneration literature suggests that is close.

§See also

• Osseointegration
• Neuroprosthetics
• Powered exoskeletons
• Brain–computer interface
• Sensory augmentation
• Limb regeneration
• Enhancement in sport
• Disability rights and enhancement

§References