Heart Rate Tracking using Wrist-Type Photoplethysmographic (PPG) Signals during Physical Exercise with Simultaneous Accelerometry (1504.04785v2)

Abstract: This paper considers the problem of casual heart rate tracking during intensive physical exercise using simultaneous 2 channel photoplethysmographic (PPG) and 3 dimensional (3D) acceleration signals recorded from wrist. This is a challenging problem because the PPG signals recorded from wrist during exercise are contaminated by strong Motion Artifacts (MAs). In this work, a novel algorithm is proposed which consists of two main steps of MA Cancellation and Spectral Analysis. The MA cancellation step cleanses the MA-contaminated PPG signals utilizing the acceleration data and the spectral analysis step estimates a higher resolution spectrum of the signal and selects the spectral peaks corresponding to HR. Experimental results on datasets recorded from 12 subjects during fast running at the peak speed of 15 km/hour showed that the proposed algorithm achieves an average absolute error of 1.25 beat per minute (BPM). These experimental results also confirm that the proposed algorithm keeps high estimation accuracies even in strong MA conditions.

• The paper introduces a novel algorithm that uses SVD-based motion artifact cancellation and iterative spectral analysis to enhance heart rate tracking.
• Empirical results demonstrate an average absolute error of 1.25 BPM during intense running, outperforming frameworks like TROIKA and JOSS.
• The method’s computational efficiency and robustness against motion artifacts underscore its potential for smart, wearable health monitoring devices.

Analysis of Heart Rate Tracking through Wrist-Type Photoplethysmographic Signals in Conjunction with Accelerometry

The paper presents a novel approach to track heart rate (HR) using wrist-based photoplethysmographic (PPG) signals combined with three-dimensional acceleration data during intense physical activities. This method addresses a significant challenge in wearable health monitoring devices, particularly due to motion artifacts (MAs) prevalent in PPG signals recorded from the wrist during exercise. Traditional HR monitoring methods using wrist PPG signals are often marred by major inaccuracies when subjects are in motion, making this research pivotal for advancing wearable technology applications, such as those in smartwatches.

Methodology Overview

The authors propose an algorithm consisting of two main components: Motion Artifact Cancellation and Spectral Analysis.

1. Motion Artifact Cancellation: The cornerstone of the proposed method is the use of Singular Value Decomposition (SVD) to process the acceleration data and generate reference motion artifact signals. This unique step overcomes the pitfalls of using acceleration data directly, which could impair adaptive filter convergence when applied indiscriminately. With SVD, the algorithm decomposes the acceleration signals into periodic components, allowing for precise MA reference signal extraction, subsequently facilitating adaptive noise cancellation in the PPG signals.
2. Spectral Analysis: This component utilizes the Iterative Method with Adaptive Thresholding (IMAT) for high-resolution spectrum estimation of the cleansed PPG signals. The spectrum peaks corresponding to HR are identified through this robust sparse reconstruction method, providing an advantage in terms of computational efficiency compared to models like FOCUSS.

Results and Comparative Performance

Empirical analysis on datasets comprising recordings of 12 subjects performing high-speed running tasks demonstrated an average absolute error of 1.25 beats per minute (BPM). This accuracy, sustained even under strong motion artifact conditions, underscores the algorithm's robustness. The experimentation clarified that improvements in estimation accuracy were achieved without the computational burdens associated with more complex models.

Notably, comparative assessments with state-of-the-art frameworks such as TROIKA and JOSS demonstrated superior performance in terms of both precision (as indicated by reduced absolute errors) and computational efficiency. The algorithm was also tested across various metrics such as Pearson Correlation and Estimation Variance, where it consistently showed high fidelity and correlation.

Implications and Future Directions

The proposed algorithm stands to make significant contributions to the design of intelligent wearable devices capable of reliable HR monitoring under real-world conditions. Its efficient handling of motion artifacts holds promise for practical deployment in consumer-grade electronics, paving the way for real-time health monitoring in fitness and medical applications.

From a theoretical perspective, this work reinforces the utility of advanced signal processing techniques like SVD and IMAT in biomedical applications, highlighting their adaptability and precision. Future research may seek to refine these methods further, potentially exploring alternative decomposition techniques or enhancing adaptive filter algorithms to support a broader range of biometric measurements beyond HR.

In summary, the paper delivers a meaningful contribution to the enhancement of HR monitoring technologies within wearable devices, enabled by innovative signal processing approaches that mitigate the challenges posed by motion artifacts inherent in wrist-based PPG measurements. As wearable technology continues to evolve, such developments are critical in broadening the scope and accuracy of non-invasive health monitoring.