Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TAU: Modeling Temporal Consistency Through Temporal Attentive U-Net for PPG Peak Detection

Chunsheng Zuo, Yu Zhao, Juntao Ye

TL;DR

The paper addresses robust PPG peak detection under motion artifacts by introducing TAU, a Temporal Attentive U-Net that fuses amplitude and temporal information through a dedicated time module and temporal embeddings. It uses distance-transform labels for smooth supervision and an encoder–decoder with multi-head attention to capture local and global context. TAU demonstrates improved peak detection, heart-rate estimation, and HRV correlation across four public datasets, with TAU-lite offering a lighter, faster variant suitable for wearables. The results underscore the value of explicit temporal modeling for reliable PPG-based cardiovascular metrics in real-world, noisy conditions.

Abstract

Photoplethysmography (PPG) sensors have been widely used in consumer wearable devices to monitor heart rates (HR) and heart rate variability (HRV). Despite the prevalence, PPG signals can be contaminated by motion artifacts induced from daily activities. Existing approaches mainly use the amplitude information to perform PPG peak detection. However, these approaches cannot accurately identify peaks, since motion artifacts may bring random and significant amplitude variations. To improve the performance of PPG peak detection, the time information can be used. Specifically, heart rates exhibit temporal consistency that consecutive heartbeat intervals in a normal person can have limited variations. To leverage the temporal consistency, we propose the Temporal Attentive U-Net, i.e., TAU, to accurately detect peaks from PPG signals. In TAU, we design a time module that encodes temporal consistency in temporal embeddings. We integrate the amplitude information with temporal embeddings using the attention mechanism to estimate peak labels. Our experimental results show that TAU outperforms eleven baselines on heart rate estimation by more than 22.4%. Our TAU model achieves the best performance across various Signal-to-Noise Ratio (SNR) levels. Moreover, we achieve Pearson correlation coefficients higher than 0.9 (p < 0.01) on estimating HRV features from low-noise-level PPG signals.

TAU: Modeling Temporal Consistency Through Temporal Attentive U-Net for PPG Peak Detection

TL;DR

The paper addresses robust PPG peak detection under motion artifacts by introducing TAU, a Temporal Attentive U-Net that fuses amplitude and temporal information through a dedicated time module and temporal embeddings. It uses distance-transform labels for smooth supervision and an encoder–decoder with multi-head attention to capture local and global context. TAU demonstrates improved peak detection, heart-rate estimation, and HRV correlation across four public datasets, with TAU-lite offering a lighter, faster variant suitable for wearables. The results underscore the value of explicit temporal modeling for reliable PPG-based cardiovascular metrics in real-world, noisy conditions.

Abstract

Photoplethysmography (PPG) sensors have been widely used in consumer wearable devices to monitor heart rates (HR) and heart rate variability (HRV). Despite the prevalence, PPG signals can be contaminated by motion artifacts induced from daily activities. Existing approaches mainly use the amplitude information to perform PPG peak detection. However, these approaches cannot accurately identify peaks, since motion artifacts may bring random and significant amplitude variations. To improve the performance of PPG peak detection, the time information can be used. Specifically, heart rates exhibit temporal consistency that consecutive heartbeat intervals in a normal person can have limited variations. To leverage the temporal consistency, we propose the Temporal Attentive U-Net, i.e., TAU, to accurately detect peaks from PPG signals. In TAU, we design a time module that encodes temporal consistency in temporal embeddings. We integrate the amplitude information with temporal embeddings using the attention mechanism to estimate peak labels. Our experimental results show that TAU outperforms eleven baselines on heart rate estimation by more than 22.4%. Our TAU model achieves the best performance across various Signal-to-Noise Ratio (SNR) levels. Moreover, we achieve Pearson correlation coefficients higher than 0.9 (p < 0.01) on estimating HRV features from low-noise-level PPG signals.

Paper Structure

This paper contains 23 sections, 7 equations, 9 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Time-Domain Signal-Processing Approaches
  4. Frequency-Domain Signal-Processing Approaches
  5. Deep Learning Approaches
  6. Heart Rate Variability Analysis
  7. Temporal Modeling in Deep Learning
  8. The Proposed Temporal Attentive U-Net Model
  9. Labeling Peaks with Distance Transform
  10. Model Architecture
  11. Encoder Module
  12. Time Module
  13. Decoder Module
  14. Prediction Module
  15. Experimental Setup
  16. ...and 8 more sections

Figures (9)

  • Figure 1: An illustration of PPG signals, systolic peaks and P-P intervals.
  • Figure 2: An illustration of distance transform labels.
  • Figure 3: The architecture of our proposed TAU model.
  • Figure 4: An illustration of peak-to-peak distance sequence and sample-to-peak distance sequence.
  • Figure 5: Signal-to-noise ratios of our datasets.
  • ...and 4 more figures