Electromyography

Electromyography (EMG) is the measurement of the electrical activity that muscles produce when they contract. When a motor neuron fires, the muscle fibres it controls generate small voltage changes called motor unit action potentials; EMG records these signals either through electrodes on the skin (surface EMG, or sEMG) or through fine needles or wires inserted into the muscle (intramuscular EMG). The technique originated in clinical neurology and rehabilitation, where it is used with nerve conduction studies to diagnose nerve and muscle disorders.

In virtual reality (VR) and augmented reality (AR), surface EMG is studied and now sold as a neural input method. Sensors worn on the forearm or wrist pick up the electrical activity of the muscles that move the fingers, letting software recognise hand gestures and intended finger movements without a camera, a touch surface, or a handheld controller. The most prominent commercial example is the Meta Neural Band, a wrist-worn sEMG device that ships with Meta Platforms' Meta Ray-Ban Display glasses and grew out of Meta's 2019 acquisition of the neural-interface startup CTRL-labs.

How it works

A muscle contracts when its motor units fire. A motor unit is one motor neuron together with the muscle fibres it controls, and each firing produces a motor unit action potential that lasts roughly 5 to 30 milliseconds. The recorded EMG signal is the summation of the action potentials from the motor units within range of the electrodes. At very low effort, below about 10 percent of a maximum voluntary contraction, individual action potentials can be told apart; at higher effort the overlapping potentials blend into the random-looking waveform characteristic of EMG.

Most of the power in a surface EMG signal lies between about 10 and 400 Hz, and the raw signal is in the microvolt range, so it must be amplified, typically by a factor of around 1,000 to 1,500, before it can be processed. Intramuscular recordings contain higher-frequency content, extending to roughly 5,000 Hz. Because the signal is non-stationary and sensitive to electrode placement, skin condition, and temperature, EMG is usually analysed over short time windows and normalised when recordings taken under different conditions are compared.

Surface versus intramuscular EMG

The two recording methods trade precision against invasiveness.

Property	Surface EMG (sEMG)	Intramuscular EMG
Sensor	Electrodes resting on the skin	Needle or fine-wire electrode inserted into the muscle
Invasiveness	Non-invasive	Invasive; applied by a trained clinician
What it measures	Combined activity of many motor units near the surface (a global measure)	Activity of individual motor units, including deep muscles
Frequency range	Roughly 10-400 Hz	Up to about 5,000 Hz
Main limitation	Cannot read deep muscles; prone to cross-talk from neighbouring muscles	Discomfort; limited to small detection volumes and low contraction levels
Typical setting	Movement analysis, sports, prosthetics, consumer interfaces	Clinical diagnosis of nerve and muscle disease

Surface electrodes are easy to apply and comfortable, which makes sEMG the practical choice for wearable and consumer devices, while intramuscular EMG remains largely confined to clinical use because it can isolate single motor units and reach muscles that surface electrodes cannot.

Clinical and historical use

EMG was developed as a medical tool and is performed together with nerve conduction studies as part of an electrodiagnostic examination of the peripheral nerves, neuromuscular junction, and muscles. Needle EMG and nerve conduction studies are used to confirm or rule out conditions that are otherwise hard to diagnose, such as peripheral neuropathy and motor neuron disease. In amyotrophic lateral sclerosis, for example, the characteristic electrodiagnostic findings are normal sensory nerve conduction with abnormal motor responses, and needle EMG of muscles including the thoracic paraspinals helps distinguish ALS from disorders that resemble it. Outside diagnosis, EMG is used in rehabilitation, biofeedback, ergonomics, and the control of powered prosthetic limbs.

Use in VR and AR

The interest in EMG for virtual reality and augmented reality comes from its ability to capture hand and finger activity without optical hand tracking. Camera-based hand tracking fails when the hands leave the field of view, in poor lighting, or when one hand occludes the other; sEMG works with the hand resting at the side, in the dark, and out of any camera's view, because it reads the muscle activity that drives finger movement rather than imaging the hand. Because the signal reflects the intent to move, a worn sensor can register a gesture at or slightly before the moment the fingers visibly move.

Myo armband

An early consumer attempt at sEMG gesture control was the Myo armband, released by the Canadian company Thalmic Labs. The Myo placed eight dry EMG electrodes around the forearm together with a nine-axis inertial measurement unit, and software classified the muscle signals into a small set of hand gestures used to control computers, presentations, and robots. The device was discontinued, but it demonstrated that a multi-channel sEMG band on the forearm could drive a gesture interface, and similar armbands remain a subject of academic gesture-recognition research.

CTRL-labs and Meta

CTRL-labs was founded in 2015 by Thomas Reardon and Patrick Kaifosh, neuroscientists who earned their PhDs at Columbia University; Reardon had earlier led the team that built Microsoft's Internet Explorer browser. The company developed a wrist-worn sEMG device meant to read the signals that motor neurons send to the hand. In September 2019 Facebook, later renamed Meta Platforms, acquired CTRL-labs for a price reported between roughly 500 million and 1 billion US dollars, and folded the team into its Reality Labs research division, where Reardon became director of neuromotor interfaces.

Facebook Reality Labs first demonstrated wrist-based sEMG research publicly in March 2021, describing it as the basis for a future input method for AR glasses. A prototype wristband was later shown alongside the Meta Orion AR glasses in 2024 as their primary input device.

Nature paper

In 2025 Meta published the underlying research in the journal Nature. The paper, "A generic non-invasive neuromotor interface for human-computer interaction" by Kaifosh and colleagues, reported a wrist sEMG band paired with machine-learning models that decode muscle activity into computer input. The work described it as the first high-bandwidth neuromotor interface that generalises across people out of the box: the models were trained on data from thousands of consenting participants, so they work for a new wearer without a per-user calibration step. Meta also released a public dataset of more than 100 hours of sEMG recordings from over 300 participants.

The reported closed-loop performance, with a person using the live system, included a median of 0.66 target acquisitions per second in a continuous cursor-navigation task, 0.88 gesture detections per second in a discrete-gesture task, and handwriting at 20.9 words per minute, the last produced by writing characters with the index finger on a surface such as a table or the wearer's leg. Meta added that personalising the handwriting model on a small amount of an individual's data could improve its accuracy by up to 16 percent.

Meta Neural Band

The first consumer product from this line is the Meta Neural Band, announced on 17 September 2025 and released on 30 September 2025. It is a wrist band carrying a 16-channel sEMG array plus an inertial measurement unit, and it recognises gestures such as tapping the thumb to the index finger, swiping the thumb along the index finger, pinching, and rotating the wrist, as well as in-air handwriting. The band is sold only as part of a 799 US dollar bundle with the Meta Ray-Ban Display glasses, which it controls in place of a touchpad or controller. Meta lists an IPX7 water-resistance rating and battery life of up to 18 hours for the band. Meta has positioned the Neural Band and the Ray-Ban Display as a step between its audio-only Ray-Ban Meta glasses and the experimental Meta Orion AR glasses.