Feasibility of Heart Rate and Respiratory Rate Estimation by Inertial Sensors Embedded in a Virtual Reality Headset

Figures

• 

  Figure 5. Example of the processing steps applied to the three GYR components (x-, y- and z-axis in red, blue and black, respectively) in a subject while sitting. After a first Butterworth filtering, the magnitude is computed, and then filtered again to result in the final signal.

• 

  Figure 1. The VR device used for the experiments, with its reference for linear and rotational accelerations indicated.

• 

  Figure 2. In the experimental protocol, Coala® Heart rate Monitor device [30] was used to obtain gold standard ECG measurement during VR signals acquisition: first, it is placed on the chest for the first measurement (ECG-Chest), and then, it is held between the thumb and the middle fingers for the second measurement (ECG-Thumb).

• 

  Figure 3. Schematic representation of HR acquisition protocol (a). The red arrows (SYNC) indicate the artifacts induced by fast head rotation on the VR acquired signals, in correspondence to each start and end of the ECG recordings; this resulted in a clear artifact visible on the y-axis of the ACC signal (b) thus allowing the automatic extraction of the portion of ACC and GYR signals (M1-Chest and M2-Thumb) corresponding in time with the gold standard ECG-measurements (ECG-Chest and ECG-Thumb). In (c) the y-axis of the GYR signal corresponding to the same subject in SIT is shown, highlighting how the SYNC artifacts could not be univocally identifiable from it.

• 

  Figure 4. Schematic representation of the protocol for RESP evaluation (a). The red arrows (SYNC) indicate the artifacts induced by fast head rotation on the VR acquired signals in correspondence to each start and end of the imposed breathing through audio listening, clearly visible on the y-axis of the ACC signal (b), thus allowing the automatic extraction of the portion of ACC and GYR signals acquired under controlled breathing (RM).

• 

  Figure 6. A schematic representation of the steps in the Adjusted-FFT Method: if the power ratio was >0.8, an additional test was performed to detect the proper frequency corresponding to HR.

• 

  Figure 7. On the left, an example of the three GYR components (in black) with the corresponding output of the applied filter superimposed (in red), and on the right, the corresponding PSD with the frequency with maximum power estimated as RESP rate highlighted.

• 

  Figure 8. The median HR (25th; 75th percentiles), expressed in bpm, and individual data (as dots) measured from the chest and the thumb with the reference Coala Heart monitor in the two consecutive acquisitions (M1-Chest and M2-Thumb) for each body posture in SIT (blue), STAND (red) and SUP (green) in the 25 subjects that allowed paired comparisons for all conditions. # p < 0.01 ECG Chest vs. ECG Thumb (Wilcoxon signed rank test); * p < 0.05 between postures (Friedman Test and Wilcoxon signed rank test).

• 

  Figure 9. The median HROculus (25th; 75th percentiles) and individual data (as dots), expressed in bpm, measured from ACC (top panel) and GYR (bottom panel) signals obtained by M1-Chest and M2-Thumb, separately, for each method and body posture in SIT (blue), STAND (red) and SUP (green) in the 25 subjects that allowed paired comparisons in all conditions. * p < 0.05 and # p < 0.01 between postures (Friedman Test and Wilcoxon signed rank test).

• 

  Figure 10. Bland–Altman analyses between the gold standard HRCoala and the mean HR measurements obtained from the GYR signal HROculus of the M1-Chest (left) and M2-Thumb (right) acquisitions while in SIT position using the Adjusted-FFT method.