EMG can capture intent before movement occurs
Aliases: electromechanical delay · surface EMG · motor intent detection
What it is
A surface electrode can record motor-unit firing before the limb has moved. The lead is not mind-reading. It is surface electromyography (sEMG): neural drive has already reached the muscle fiber, while mechanical contraction has not yet produced a visible displacement. The system is capturing an electrical event at the periphery of motor output, not an abstract cortical intention.
Why it happens
A motor command reaches the neuromuscular junction; the fiber membrane depolarizes and produces an action potential that can be recorded on the skin. Force still waits on calcium release and cross-bridge cycling. That electromechanical delay (EMD) is typically about 30–100 ms and lengthens with load, cold, and longer muscles. Cameras and inertial sensors must wait for displacement to clear a noise floor; EMG has a waveform before that. The “head start” is therefore excitation–contraction coupling, not thought decoding. Envelope detection still needs an integration window, so some of the lead handed to the interface is spent there.
Studying it
A common setup records sEMG together with acceleration or joint angle and compares EMG onset with kinematic onset. Independent variables include contraction type (isometric / isotonic), load, and preprocessing (high-pass, rectification, envelope window). Dependent measures include onset lag, false triggers, and delay from the electrical event to an interface response. Gesture classifiers use time-domain features such as RMS, mean absolute value, and zero-crossing rate; information-transfer figures must state vocabulary size and inter-trial interval. Gel electrodes and a seated pose in the lab overstate the lead that dry consumer electrodes keep while walking.
Where it stops holding
EMD is not a fixed head start. Fatigue changes motor-unit recruitment; antagonist co-contraction makes onset detectors wobble. The electrical-to-mechanical mapping in amputation, spasticity, or peripheral nerve injury is not the healthy-participant mapping, so laboratory milliseconds do not transfer. Radio latency, classification windows, and firmware debounce can erase a lab finding that EMG beat a button by tens of milliseconds. Treating EMG as a brain–computer interface parks the ill-posed cortical inverse problem on a single muscle.
Applying it
- Use EMG for imminent action: preload, prerender, or mode switch—not as a replacement for already-fast touch clicks.
- Give onset detection a reproducible noise threshold and count false triggers while walking, holding objects, and talking.
- Cross-check with inertia: the interface should not fire when the limb is still and only postural EMG is present.
- Verify by aligning the EMG envelope, inertial traces, and UI logs, then separately scoring “UI before visible motion” and “trigger with no motion.”