Speaker
Aim & Purpose
This study aimed to investigate the associations of pulse-timing variability and wrist motion with heart-rate estimation error in real world wrist-PPG recordings. By studying these factors, the study seeks to better understand the limitations of wearable PPG and identify signal characteristics associated with reduced measurement reliability.
Discussion (methods + interpretation)
Wrist PPG and three-axis accelerometer recordings from the IEEEPPG dataset were analysed. The dataset contained 3,096 recordings sampled at 125 Hz, divided into 1768 TRAIN and 1328 TEST recordings, each accompanied by reference ECG-based heart rate. PPG peaks were detected to calculate inter-beat intervals (IBIs) and their coefficient of variation. Heart rate was also estimated using Welch’s power spectral density (PSD) method. Accelerometer signals were processed to quantify dynamic wrist motion. Spearman rank correlation was used to examine associations between pulse-timing variability, motion, and absolute heart-rate estimation error.
Background
Photoplethysmography is a non-invasive technique used in wearable devices to monitor heart rate. It works by detecting blood volume changes in tissue. However, wrist movement and changes in pulse-wave morphology can interfere with signal quality and peak detection, reducing estimation accuracy. Understanding such sources of error is crucial for improving the reliability of wearable heart-rate monitoring during everyday activities.
Result
Pulse-timing variability showed a strong positive association with absolute heart-rate estimation error in TRAIN recordings (Spearman ρ = 0.694, p < 0.001). Wrist motion showed a weaker positive association with estimation error in TRAIN data (ρ = 0.227, p < 0.001), which persisted in held-out TEST recordings (ρ = 0.139, p < 0.001). These findings suggest that both pulse-detection reliability and wrist movement are relevant to the accuracy of wearable PPG-based heart-rate estimation.
| Email address | nimrasoomro98@gmail.com |
|---|---|
| Phone number | 03062771960 |