Wearable and Invisible ECG Quality and Usability Assessment in Cockpit Monitoring

Figures

• 

  Figure 1. (a) Experimental setup with a volunteer in the Airbus A320 flight simulator. (b) Sensing modalities: (1) commercial pre-gelled electrodes used for wECG; (2) custom 3D-printed conductive electrodes externally mounted on the sidestick and power quadrant for iECG sensing.

• 

  Figure 2. Overview of the task-specific ECG usability framework. Raw wECG and iECG signals are converted into physical units, preprocessed, and segmented into analysis windows. Four evidence layers are then extracted, namely, contact availability, R-peak reliability, morphology evidence, and SQI support. The resulting sub-scores are fused into a global usability score and mapped to five task-specific ECG usability classes.

• 

  Figure 3. Representative iECG windows of S1 illustrating the contact states. The dashed line at zero marks the baseline reference.

• 

  Figure 4. Beat-aligned ECG morphology for wECG and iECG. (a) Beats grouped according to dSQI category. (b) Beats grouped according to the final usability class assigned by the proposed framework. Solid lines represent the mean normalized waveform, shaded areas represent inter-beat variability, and n indicates the number of aligned beats in each group. The vertical gray dashed line marks the R-peak reference (0 ms).

• 

  Figure 5. Window-level concordance between wECG and iECG. (a) Concordance matrix between final usability classes for temporally paired windows. (b) SQI-level concordance matrices between wECG and iECG for dSQI, SNR, kSQI, pSQI, and bSQI. Cell colors indicate the iECG usability class in panel (a) and the iECG SQI category in panel (b), while outlined diagonal cells denote concordant classifications between wECG and iECG.

• 

  Figure 6. Single-participant example showing simultaneous wECG and iECG signals, HR estimates, and window-level usability classes.The legend reports the complete U1–U5 color coding of the usability framework for consistency.

• 

  Figure 7. Poincaré plots of consecutive RR intervals for wECG and iECG, colored by usability class. Ellipses represent class-specific RR dispersion. U5 windows are excluded.

• 

  Figure 8. HR agreement between iECG and wECG according to the proposed usability framework. (a) Left: MAEs for concordant and discordant paired windows. Right: iECG-derived HR versus wECG-derived HR, with concordant and discordant windows highlighted separately. (b) Bland–Altman analysis for iECG windows classified as U1–U3.