Validation of the EmotiBit wearable sensor for heart-based measures under varying workload conditions

Figures

• 

  Figure 1. Light wavelengths of EmotiBit PPG sensor. The absorption spectra of oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (Hb) represented by the molar extinction coefficient (cm−1/M) on a logarithmic scale is plotted against wavelength (nm) with highlighted bands corresponding to specific light wavelengths used in the MAX30101 sensor (values derived from SparkFun Photodetector (MAX30101) Hookup Guide - SparkFun Learn, 2025). The highlighted bands include the wavelengths measured of the green light spectrum (530–545 nm), the red light spectrum (650–670 nm), and the infrared light spectrum (870–900 nm) built into the sensor. Reproduced from Prahl (1999).

• 

  Figure 2. Experimental pocedure. (A) The timeline of the experimental procedure is illustrated with time estimates for each phase. The experimental block consisted of an operation span task with an easy and hard workload condition counterbalanced in order across participants. Note that the adapted Wiener Matrices Test had a fixed duration while the other time periods were dependent on the individual speed of completion. (B) The experimental blocks are illustrated for both workload conditions. During the visual OSPAN task, participants performed mental calculations and delayed recalls and responded with button presses (left or right) counterbalanced across participants. This was interjected with three types of questionnaires to assess current mental load, stress, and emotional state. An auditory implicit oddball task was presented and implicitly attended to with counterbalanced frequencies for oddball and frequent sounds and jittered inter-stimulus intervals. BFI10, Big Five Inventory-10 (Rammstedt et al., 2017); ISI, inter-stimulus interval; MDBF, German Multidimensional Mood State Questionnaire (Hinz et al., 2012); NASA-TLX, NASA Task Load Index (Hart and Staveland, 1988); OSPAN,…

• 

  Figure 3. Picture of the sensor positioning at the arm. The EmotiBit device was placed on the upper arm and the PsychoBit EDA sensors were adhered to the left palm. Additionally, the PsychoBit pulse sensor was attached to the index finger and a consumer wearable (Fitbit Sense 2) was positioned above the wrist, albeit not reported in the present study.

• 

  Figure 4. Processing pipelines for CVA and EDA signals. The processing pipelines are depicted for ECG (blue), PPG (red), and EDA (green) data. For data cleaning, signal decomposition of the EDA, and peak detection, various methods were tested. ECG, electrocardiography; EDA, electrodermal activity; PPG, photoplethysmography; RMSSD, root mean square difference of the successive differences; SCL, skin conductance level; SCR, skin conductance response; SD, standard deviation.

• 

  Figure 5. Device comparison of CVA parameters of interest. (A) Bland-Altman plots for the parameters' comparison of the CVA signals in terms of the log-transformed mean HR, the SD of the RR/PP intervals, and the RMSSD. Colored dots represent individual participants during the easy condition (left) and the hard condition (right). Each color corresponds to one participant. The x-axis corresponds to the average of the two measures for one parameter, and the y-axis shows the difference between the two measures. The theoretical difference of zero is marked as a gray dotted line. The observed mean difference is plotted as a dashed blue line with the standard deviations marked as blue areas. The orange lines represent the observed 95% confidence interval limits. The computed BAr is indicated within each plot. (B) Violin plots of z-scored parameters of interest per condition for ECG (blue) and EmotiBit PPG (red). The median (solid lines) and 1st and 3rd quartiles (dashed lines) for each measured parameter of interest are indicated per device. Bar, Bland-Altman ratio; Δ, Mean difference; ECG, Electrocardiography; HR, Heart rate; PPG, Photoplethysmography; RMSSD: Root mean square of successive…

• 

  Figure 6. Highest cross-correlation coefficients across subjects. The highest cross-correlation of raw (top) and processed (bottom) EDA signals of is plotted as histogram devices between −8 and +8 sample time lags (corresponding to −0.53 to 0.53 s with a sampling rate of 15 Hz) is plotted for each run divided by easy (blue) and hard (red) condition of the workload task. The mean cross-correlations are indicated as dashed lines.

• 

  Figure 7. Device comparison of EDA parameters of interest. (A) Bland-Altman plots for the parameters' comparison of the EDA signals in terms of the log-transformed mean SCL, the number of SCRs per minute, and the mean amplitude of SCRs. Colored dots represent individual participants during the easy condition (left), and the hard condition (right). Each color corresponds to one participant. The x-axis corresponds to the average of the two measures for one parameter, and the y-axis shows the difference between the two measures. The theoretical difference of zero is marked as a gray dotted line. The observed mean difference is plotted as a dashed blue line with the standard deviations marked as blue areas. The orange lines represent the observed 95% confidence interval limits. The computed BAr is indicated within each plot. (B) Violin plots of z-scored parameters of interest per condition for ECG (blue) and EmotiBit PPG (red). The median (solid lines) and 1st and 3rd quartiles (dashed lines) for each measured parameter of interest are indicated per device. Abbeviations: Bar, Bland-Altman ratio; Δ, Mean difference; ECG, Electrocardiography; HR, Heart rate; PPG, Photoplethysmography; RMSSD,…

• 

  Figure 8. Example of peak detection in PPG signal streams. The figure illustrates the effect of averaging the three light wavelength signals of a PPG (PG, PR, PI) on the peak detection comparability to an ECG in the case of low variance between the signal streams (left) compared to high variance in the signal streams (right). In the left panel, an example of consistent peak-to-peak intervals across signals is highlighted with arrows. The right panel shows inconsistent peak-to-peak intervals across signals. ECG, Electrocardiography; PG: green PPG signal; PI, infrared PPG signal; PP, P-peak interval PPG, photoplethysmography; PR, red PPG signal; RR, R-peak interval.