TROIKA: A General Framework for Heart Rate Monitoring Using Wrist-Type Photoplethysmographic Signals During Intensive Physical Exercise

Abstract: Heart rate monitoring using wrist-type photoplethysmographic (PPG) signals during subjects' intensive exercise is a difficult problem, since the signals are contaminated by extremely strong motion artifacts caused by subjects' hand movements. So far few works have studied this problem. In this work, a general framework, termed TROIKA, is proposed, which consists of signal decomposiTion for denoising, sparse signal RecOnstructIon for high-resolution spectrum estimation, and spectral peaK trAcking with verification. The TROIKA framework has high estimation accuracy and is robust to strong motion artifacts. Many variants can be straightforwardly derived from this framework. Experimental results on datasets recorded from 12 subjects during fast running at the peak speed of 15 km/hour showed that the average absolute error of heart rate estimation was 2.34 beat per minute (BPM), and the Pearson correlation between the estimates and the ground-truth of heart rate was 0.992. This framework is of great values to wearable devices such as smart-watches which use PPG signals to monitor heart rate for fitness.

• The paper presents TROIKA, a framework that integrates signal decomposition, SSR, and spectral peak tracking to enable accurate HR monitoring from wrist PPG signals during intensive exercise.
• It employs SSA to reduce motion artifacts and SSR with FOCUSS for high-resolution spectral analysis, achieving an average error of 2.34 BPM and a 0.992 Pearson correlation.
• The method's robustness at lower sampling frequencies enhances its applicability for power-efficient, consumer-grade wearables in rigorous physical activities.

"TROIKA: A General Framework for Heart Rate Monitoring Using Wrist-Type Photoplethysmographic Signals During Intensive Physical Exercise" (1409.5181)

Introduction

The paper introduces TROIKA, a robust framework designed to accurately monitor heart rate (HR) using wrist-type photoplethysmographic (PPG) signals during intensive physical exercise. This has emerging significance, especially for wearable devices like smartwatches that face challenges due to motion artifacts (MA) resulting from hand movements.

Theoretical Framework

TROIKA comprises three primary components:

• Signal Decomposition: Using singular spectrum analysis (SSA), this part aims to partially eliminate motion artifacts. SSA decomposes a signal into oscillatory components and noise, allowing the removal of noise components to reconstruct a cleaner signal. This preprocessing is critical to sparsify the PPG signal's spectrum, making subsequent analysis more effective.
• Sparse Signal Reconstruction (SSR): TROIKA employs SSR for high-resolution spectral analysis. By framing the problem as an optimization challenge, TROIKA leverages SSR with a matrix $\mathbf{\Phi}$ representing basis functions in which the signal is sparse. Here, SSR bypasses the requirements of model order selection that plague traditional spectrum estimation models. The Fast Orthogonal Search (FOCUSS) algorithm is used due to its robustness with highly correlational basis matrices.
• Spectral Peak Tracking: This module leverages a multistage process – including peak selection and verification – to track HR in real-time. Techniques like peak harmonics validation and temporal consistency checks ensure accurate peak identification amidst spectral complexity caused by MA.

Experimental Validation

The framework was validated using data from 12 male subjects who underwent sessions of walking and running at variable speeds on treadmills. The TROIKA system was tasked with continuously estimating HR using both PPG and acceleration signals recorded at 125 Hz. The average absolute error across subjects was 2.34 BPM, and the Pearson correlation coefficient between estimated and true HR pointed to high accuracy (0.992).

System Robustness

Testing involved parameter sensitivity analyses to ensure TROIKA's efficacy across different conditions. Testing also evaluated performance under scenarios where any of the core algorithms were substituted or omitted, underscoring the indispensability of each module. For instance, omitting the spectral peak verification step led to substantial performance degradation, highlighting the importance of the holistic framework.

Implications

The high fidelity of the TROIKA framework provides a promising solution for integrating accurate HR monitoring into consumer wearables without requiring large computational resources or multiple PPG channels. Its ability to maintain performance with reduced sampling frequency (down to 25 Hz) significantly enhances the practical applicability and power efficiency of wearable devices.

Conclusion

TROIKA demonstrates the capability to overcome significant challenges in HR monitoring during intense activities, utilizing advanced signal processing techniques to offset noise introduced by motion artifacts. Its methodology can be customized to a variety of wearable technologies, offering a balance between accuracy and operational efficiency. Future work could explore enhancements to SSR algorithms or explore alternate signal decomposition methods to further improve the framework's adaptability and performance under varying physical dynamics.

In summary, TROIKA addresses a critical need in the wearable technology domain, ensuring reliable and accurate heart rate monitoring during physical exercise. The innovative combination of signal decomposition, SSR, and spectral tracking constitutes a flexible and high-performing framework applicable across various wearable platforms.