HSF-IBI: A Universal Framework for Extracting Inter-Beat Interval from Heterogeneous Unobtrusive Sensors

Zhongrui Bai, Pang Wu, Fanglin Geng, Hao Zhang, Xianxiang Chen, Lidong Du, Peng Wang, Xiaoran Li, Zhen Fang, Yirong WuBioengineering 2024 · DOI

TL;DR

This paper introduces HSF-IBI, a novel framework for accurately extracting inter-beat intervals (IBIs) from various unobtrusive sensors by fusing their data. It achieves superior accuracy and detection rates compared to existing methods, paving the way for more robust and reliable remote health monitoring.

What Problem Does This Paper Address?

Monitoring vital signs like heart rate unobtrusively is a growing area in healthcare, offering comfort and convenience over traditional methods. A key metric derived from this is the Inter-Beat Interval (IBI), which provides insights into cardiac and neurological function. However, existing unobtrusive sensors (like those measuring body vibrations or radar signals) often suffer from noise and signal variability, making precise and reliable IBI extraction a significant challenge. This variability makes it difficult to develop a universal method that works across different sensor types and signal morphologies. The paper aims to bridge this gap by proposing a framework that can effectively fuse data from heterogeneous unobtrusive sensors to overcome individual sensor limitations and achieve more accurate IBI estimation.

Methods

The core of the HSF-IBI framework is a multi-stage process designed for robustness and adaptability:

  • Harmonic Summation (HarSum-HR): This novel approach is used to estimate the mean heart rate. It leverages harmonic summation, a technique adapted from speech processing, to analyze the frequency spectrum of the cardiac vibration signals. This provides a stable reference for subsequent steps.
  • Optimal Band Selection (OBS): Before extracting heartbeats, the framework intelligently selects the optimal frequency band for each sensor's signal. This is crucial because different sensors and body positions can shift the relevant heartbeat signal components to different frequency ranges. OBS helps to isolate the most informative parts of the signal.
  • Template Group Extraction (BSAS): Instead of relying on a single template for heartbeat detection, HSF-IBI uses a group of templates extracted using the Basic Sequential Algorithmic Scheme (BSAS). This accounts for variations in heartbeat morphology, especially those influenced by respiration.
  • Template Matching (TM) and Signal Quality Index (SQI): The extracted templates are then used for matching against the processed signals. Crucially, the algorithm uses a Signal Quality Index (SQI) for each detected heartbeat to assess its reliability. This allows for a beat-to-beat fusion strategy.
  • Heterogeneous Sensor Fusion: The framework combines the IBIs from multiple sensors by selecting the one with the highest SQI for each heartbeat, provided it meets a certain quality threshold. This fusion strategy leverages the strengths of different sensors and mitigates the weaknesses of individual ones.

Datasets

The proposed HSF-IBI framework was evaluated on two datasets:

  • Dataset A: A self-collected dataset comprising data from 19 healthy participants. This dataset included signals from radar and piezoelectric ceramic (PEC) sensors, collected during both cycling and supine rest.
  • Dataset B: An open-source dataset from Carlson [40], containing data from 34 healthy participants. This dataset utilized eight-channel bed-based ballistocardiograms (BCGs) acquired using electromechanical films (EMFi) and load cells (LC).

Both datasets involved heterogeneous unobtrusive sensors, allowing for the evaluation of the framework's ability to integrate diverse signal types.

Key Insights and Ideas

  • Fusion is Key for Robustness: The paper strongly demonstrates that combining data from different types of unobtrusive sensors significantly improves both the accuracy and coverage of IBI estimation. This is not just about having more data, but about leveraging the complementary nature of different sensor sensitivities to noise and signal characteristics.
  • Morphology-Independent Approach: The HSF-IBI framework is designed to be morphology-independent. This is a critical advantage because unobtrusive sensor signals can vary greatly in shape depending on the sensor, placement, and even the subject's posture. This adaptability is a major step towards a universal solution.
  • HarSum-HR for Stable Heart Rate Estimation: The novel HarSum-HR method provides a stable and adaptable way to estimate mean heart rate, even when signal morphology changes due to postural shifts. This is crucial for guiding subsequent IBI extraction steps.
  • Optimal Band Selection Matters: Identifying the optimal frequency band for each sensor's signal is vital. The OBS module helps to isolate the most informative components of the cardiac vibration signals, leading to more accurate IBI detection.

Key Results

The HSF-IBI framework demonstrated significant improvements:

  • Detection Rate Enhancement: The framework achieved detection rate enhancements of 6.25% on Dataset A and 5.21% on Dataset B compared to single-sensor or single-type sensor fusion methods.
  • Superior Accuracy: The proposed framework achieved superior accuracy over existing IBI extraction methods, with mean absolute errors (MAEs) of 5.25 ms on Dataset A and 4.56 ms on Dataset B.
  • High Coverage: The heterogeneous sensor fusion scenario significantly improved coverage, reaching 93.22% on Dataset A and 97.83% on Dataset B.
  • State-of-the-Art Performance: The paper claims the proposed framework achieves state-of-the-art accuracy for unobtrusive IBI estimation.

Limitations and Future Work

  • Dataset Size and Diversity: The authors acknowledge that the datasets, while incorporating multiple sensors and participants, still have a relatively small number of participants. Future work should focus on validating the method with larger and more diverse participant pools, including those with heart conditions.
  • Computational Demands: While the algorithm is designed to be computationally efficient, integrating multiple channels inherently increases computational load. The authors suggest that real-time computation is feasible, but further optimization might be needed for resource-constrained devices.
  • Abnormal Heartbeats: The current framework might need modifications to effectively handle abnormal heartbeats, which were excluded from some analyses. Tailoring the method for patients with heart conditions is a key future direction.

Summary updated

Abstract

Heartbeat inter-beat interval (IBI) extraction is a crucial technology for unobtrusive vital sign monitoring, yet its precision and robustness remain challenging. A promising approach is fusing heartbeat signals from different types of unobtrusive sensors. This paper introduces HSF-IBI, a novel and universal framework for unobtrusive IBI extraction using heterogeneous sensor fusion. Specifically, harmonic summation (HarSum) is employed for calculating the average heart rate, which in turn guides the selection of the optimal band selection (OBS), the basic sequential algorithmic scheme (BSAS)-based template group extraction, and the template matching (TM) procedure. The optimal IBIs are determined by evaluating the signal quality index (SQI) for each heartbeat. The algorithm is morphology-independent and can be adapted to different sensors. The proposed algorithm framework is evaluated on a self-collected dataset including 19 healthy participants and an open-source dataset including 34 healthy participants, both containing heterogeneous sensors. The experimental results demonstrate that (1) the proposed framework successfully integrates data from heterogeneous sensors, leading to detection rate enhancements of 6.25 % and 5.21 % on two datasets, and (2) the proposed framework achieves superior accuracy over existing IBI extraction methods, with mean absolute errors (MAEs) of 5.25 ms and 4.56 ms on two datasets.
BibTeX
@inproceedings{bai2024hsf,
  author = {Zhongrui Bai and Pang Wu and Fanglin Geng and Hao Zhang and Xianxiang Chen and Lidong Du and Peng Wang and Xiaoran Li and Zhen Fang and Yirong Wu},
  title = {{HSF-IBI: A Universal Framework for Extracting Inter-Beat Interval from Heterogeneous Unobtrusive Sensors}},
  booktitle = {Bioengineering},
  year = {2024},
  doi = {10.3390/bioengineering11121219},
}

Loading related content...

Figure