Effects of sensor geometry, placement, and cycle detection on wearable respiration monitoring with textile printed strain sensors

Abstract

Wearable respiration monitors often struggle with motion artifacts and variable skin–sensor coupling. We present an elastic chest belt with digitally printed piezoresistive strain gauges and a multi-channel bridge front-end, and we evaluate how sensor geometry, placement, and simple respiration cycle detectors affect minute-scale respiration-rate (RR) accuracy. Three printed sensor layouts (circular and serpentine/“linear”) were placed at three thoracic locations (left, center, right); for the serpentine layout, we also tested horizontal vs. vertical orientations. RR from (i) a time-domain peaks-and-valleys (PV) detector and (ii) a power spectral density (PSD) method were compared to a thermistor reference across 4-min trials in healthy volunteers (n = 3), analyzed per minute. PV achieved a larger share of minutes within ± 2 RPM than PSD (≈ 70% vs. ≈37%) and within ± 1 RPM (≈ 44% vs. ≈33%). Geometry and placement mattered: a vertical serpentine on the lateral thorax showed the highest minute-level agreement within this pilot cohort and under quiet-breathing conditions (left-lateral vertical: 100% of minutes within ± 2 RPM; right-lateral vertical/horizontal: 91.7%), while circular sensors showed their highest accuracy on the left thorax (83.3% within ± 2 RPM). These pilot findings suggest preliminary design guidance for e-textile RR monitors and indicate that PV detection may be a practical option for minute-scale RR estimation under the tested conditions.