Effects of Ballistocardiogram Peak Detection Jitters on the Quality of Heart Rate Variability Features: A Simulation-Based Case Study in the Context of Sleep Staging

Figures

• 

  Figure 1. Heartbeat interval estimates from simultaneous ECG and BCG recordings. An R–R interval (RRI) represents the time between consecutive R peaks—the prominent peaks of an ECG that correspond to the onset of ventricular depolarization events. Similarly, a J–J interval (JJI) represents the time between consecutive BCG J peaks, where J peaks are the prominent peaks of a BCG, believed to be generated by blood ejection events at the aortic artery. Variations in RJIs due to physiology are also indicated in this figure, labeled as RJIn. Although RJI1, RJI2, and RJI3 are visually similar, they are numerically different. The signal amplitudes are normalized relative to their respective ranges across the entirety of each recording.

• 

  Figure 2. An example of perturbations in J–peak time estimates due to peak detection algorithm performance.

• 

  Figure 3. RJI data and a theoretical Gaussian fit based on mean and standard deviation of the RJI data.

• 

  Figure 4. The process of adding jitter to the R-peak locations and recording the resulting RRIs.

• 

  Figure 5. SVM classification baseline performance due to all HRV features for micro-labeled sleep stages.

• 

  Figure 6. SVM classification error due to all HRV features for micro-labeled sleep stages. Light gray lines represent classification errors for each participant. The dashed, bold black line is the ensemble average, and the dotted bold lines show mean +/− 1.96 standard deviations.

• 

  Figure 7. SVM classification error due to all HRV features for macro-labeled sleep stages. Light gray lines represent classification errors for each participant. The dashed, bold black line is the ensemble average, and the dotted bold lines show mean +/− 1.96 standard deviations.