Electrode placement and skin condition affect signal quality

What it is

The same muscle and the same contraction can look like a different gesture if the electrodes move a centimetre or the stratum corneum has just started sweating. Whether EMG is usable as input is decided first at the electrode–skin interface: placement, spacing, and impedance, not the depth of the later model.

Why it happens

Bipolar surface EMG records a potential difference between two sites. Motor-unit action potentials propagate from the innervation zone toward both tendon ends. Electrodes that straddle that zone see nearly matching waveforms and cancel in the difference, collapsing SNR. Wider spacing recruits more motor units and also more crosstalk from neighbours, especially over thick subcutaneous fat. Water in the stratum corneum, sebum, hair, and gel versus dry contact set impedance: high impedance couples mains hum and cable micromotion into the channel, looking like a contraction. Dry electrodes are higher and less stable in impedance, so they punish placement error harder.

Studying it

A standard probe is to translate electrodes systematically on either side of the innervation zone, or to vary spacing, while measuring amplitude at maximal voluntary contraction, median frequency, and coherence with a neighbouring muscle. Skin preparation (alcohol, abrasion, sweat) is a separate factor. Outcomes include SNR, how classification accuracy falls with placement error, and impedance spectra. SENIAM-style placement guides are a reproducible laboratory start, not a guarantee for a consumer wristband. Cross-day donning must treat “remove and reattach” as a condition, or placement error is billed to the algorithm.

Where it stops holding

Anatomical landmarks are easy on a lean forearm and drift on adipose tissue, oedema, children, or residual limbs. Laboratory SNR on gel electrodes does not represent all-day dry textile contacts. Sweat can first drop impedance and enlarge the signal, then dilute electrolyte and lift electrode edges—quality can rise then fall inside one recording. Writing “place on the forearm extensors” as a fixed recipe, while ignoring that the innervation zone varies across people, yields a near-zero differential in some users.

Applying it

• Place electrodes on the target muscle belly rather than on the look of the strap, and avoid the innervation zone; use arrays in shipping hardware to buy tolerance of placement error.
• Monitor contact impedance or common-mode energy in firmware and stop emitting gesture labels when contact is lost.
• Instruct users to clean skin and avoid wounds or heavy hair; accept first use on wet hands or after lotion as a separate test.
• Verify by having users self-don three times from the instructions and comparing amplitude and confusion, not a single experimenter-placed session.

Related

• Same group: C9.01.1 EMG can capture intent before movement occurs · C9.01.3 Tiny movements reduce fatigue but also reduce discoverability
• Adjacent: C9.09 Wearable Fit and Signal Quality · C9.10 Individual Calibration and Baseline Drift
• Search: electrode-skin impedance · innervation zone · EMG crosstalk