Loose or misplaced wear systematically shifts the measurement baseline

What it is

A strap that slides toward the ulnar styloid, a headband that misses its 10–20 target, a chest belt that rides to one side: the measurement is no longer the same tissue. The offset is systematic. The new site has its own DC level and waveform shape, stably unlike the calibration take. That is not the person’s physiological baseline drifting with fatigue—the part moved house.

Why it happens

Optical heart rate sees different perfusion and tendon optics on the dorsal wrist, the side, and near the palm; resting rate can differ by several beats, and variability more so. EDA electrodes leaving the palm for the dorsal wrist drop tonic level as a block and shrink phasic responses. EMG electrodes sliding off the belly decay amplitude by volume conduction; the classifier still uses the old template, which is a different channel. Once misplaced, the bias is approximately constant until the next donning, so adaptive algorithms write it in as a “new baseline” and mis-count every later relative change. Looseness also adds intermittent contact: the baseline jumps between two sites, harder to clear with a simple high-pass than a fixed misplacement.

Studying it

Mark wear position, translate by millimetres to centimetres on purpose, and compare rest level and task-evoked amplitude. Factors: direction of shift, tightness. Outcomes: DC offset, template correlation, accuracy drop of a classifier left on the old calibration. Treat re-donning as a condition: how wide is position spread when users follow the manual a second time. Optics can use a perfusion map to explain site differences; EEG should refer to cap electrode labels.

Where it stops holding

Some quantities are insensitive to small misplacement (step count); some are extremely sensitive (bipolar EMG). A well-elastic headband may recentre during motion, making offset temporary. In children and oedema, anatomical landmarks move during the day, and misplacement cannot be separated from body change. If the system never individually calibrates, misplacement shows up as absolute error, not as a shift “relative to a calibration point.”

Applying it

• Constrain the wearable region with shell shape, recesses, or printed alignment marks to cut unconstrained slide.
• When DC jumps relative to the calibration point, demand a re-don or void that day’s relative baseline; do not treat the new DC as a physiological event.
• Teach with a “correct site / common slip” pair of pictures for that hardware.
• Verify by running a task after donning, then loosening one notch: if relative metrics translate as a block and go unmarked, mechanical baseline shift was not recognized.

Related

• Same group: C9.09.1 How well an electrode or sensor seats on skin directly sets signal-to-noise ratio · C9.09.2 Sweat, hair, and changing skin alter contact impedance during wear · C9.09.4 Users can rarely judge fit themselves; the system needs to show signal quality
• Adjacent: C9.10 Individual Calibration and Baseline Drift · C8.10 Gaze Drift and Recalibration
• Search: sensor misplacement · baseline offset · wearable slip