Accelerometer‐based heart rate adjustment for ambulatory stress research

Figures

• 

  FIGURE 1. Flowchart of the data processing pipeline showcasing the inclusion of participants, the algorithm for posture and activity type recognition, the processing steps for the heart rate data, the statistical testing of the four different heart rate prediction models, and the further analysis with the best model.

• 

  FIGURE 2. Confusion matrix containing the results of the activity recognition algorithm. All values are proportional to the actual occurrence (number of positives in the ground truth) of the category.

• 

  FIGURE 3. Adjusted R 2 and RMSE for each of the four models.

• 

  FIGURE 4. Overview of the observed and predicted heart rate averaged over all participants for both study protocols. The top plot shows the first study protocol, while the bottom plot shows the second study protocol. Vertical colored bars indicate the different categories present in the study protocol. The color coding is clarified in the external legend to the right of the plot.

• 

  FIGURE 5. Overview of the additional heart rate, as well as the observed heart rate reactivity, averaged over all participants for both study protocols. The top plot shows the first study protocol, while the bottom plot shows the second study protocol. Vertical colored bars indicate the different categories present in the study protocol. The color coding is clarified in the external legend to the right of the plot.

• 

  FIGURE 6. Scatter plot with aHR on the x axis and ΔHR on the y axis. The blue dots represent average ΔHR and aHR per participant in the specific conditions. Top four plots represent physical activity conditions, and have a wider scale than the mental stress conditions in the bottom six plots. The number of participants per condition is indicated in the title above all separate plots, as well as the correlation coefficient. The diagonal line represents the identity line, and the larger black dot represents the center of mass.

• 

  FIGURE 7. Manipulation check displaying the average and standard deviation per item of negative affect and positive affect.

• 

  FIGURE 8. Confusion matrix of the within‐subject model validation analysis. All values are calculated within participants and subsequently transformed to be proportional to the actual occurrence (number of positives in the ground truth) of the category. Then, the values in the confusion matrix are averaged across participants and presented in combination with the standard deviation across participants between brackets. The left panel indicates the classification approach executed with additional heart rate, whereas the right panel indicates the classification approach executed with the observed heart rate.