Photoplethysmography-Based Heart Rate Monitoring in Physical Activities via Joint Sparse Spectrum Reconstruction (1503.00688v2)

Abstract: Goal: A new method for heart rate monitoring using photoplethysmography (PPG) during physical activities is proposed. Methods: It jointly estimates spectra of PPG signals and simultaneous acceleration signals, utilizing the multiple measurement vector model in sparse signal recovery. Due to a common sparsity constraint on spectral coefficients, the method can easily identify and remove spectral peaks of motion artifact (MA) in PPG spectra. Thus, it does not need any extra signal processing modular to remove MA as in some other algorithms. Furthermore, seeking spectral peaks associated with heart rate is simplified. Results: Experimental results on 12 PPG datasets sampled at 25 Hz and recorded during subjects' fast running showed that it had high performance. The average absolute estimation error was 1.28 beat per minute and the standard deviation was 2.61 beat per minute. Conclusion and Significance: These results show that the method has great potential to be used for PPG-based heart rate monitoring in wearable devices for fitness tracking and health monitoring.

Photoplethysmography-Based Heart Rate Monitoring via Joint Sparse Spectrum Reconstruction

This essay provides a detailed overview of a paper examining heart rate (HR) monitoring utilizing photoplethysmography (PPG) signals influenced by motion artifact (MA). The research introduces a novel methodology for estimating HR during physical activities, leveraging joint sparse spectrum reconstruction techniques. This approach, termed JOint Sparse Spectrum reconstruction (JOSS), addresses complications associated with MA contamination in PPG signals, offering significant improvements over existing methods such as TROIKA.

Methodology

The authors propose an innovative HR monitoring method based on the multiple measurement vector (MMV) model, a model identified for its capability to simultaneously process multiple signals sharing common sparse structures. JOSS distinctly diverges from its predecessors by utilizing this MMV model to estimate spectra of both PPG and acceleration signals in unison. This approach contrasts with single measurement vector (SMV) models, like that used in TROIKA, and offers enhanced performance by exploiting common sparsity constraints innate to the joint signals.

The MMV model effectively aligns spectral peaks of MA in PPG spectra with corresponding peaks in acceleration spectra, thus facilitating accurate MA removal through spectral subtraction. This operation significantly reduces the need for complex signal decomposition stages present in other algorithms. By simplifying the HR estimation process, JOSS demonstrates suitability for practical applications in hardware-constrained environments such as wearable devices.

Results

The evaluation conducted using 12 PPG datasets reveals an average absolute estimation error of 1.28 beats per minute (BPM) with a standard deviation of 2.61 BPM, a significant improvement over TROIKA's 2.42 BPM average error. The experimental setup involved processing PPG signals downsampled to 25 Hz, further demonstrating JOSS's robustness to lower sampling rates—a critical feature for energy-efficient wearable health monitoring devices.

Practical and Theoretical Implications

Practically, JOSS offers a substantial step forward by augmenting the usability of PPG-based HR monitoring in wearable technology, particularly under conditions of strong MA and low sampling rates. Theoretically, it underscores the potential of exploiting joint sparsity structures in multichannel signal processing, which could stimulate further advancements in sparse signal recovery methodologies.

Future Developments

While JOSS enables effective HR monitoring with reduced computational complexity, future research could investigate more sophisticated sparse recovery models exploiting other existing structures in signal spectra. Moreover, the adoption of gridless joint spectral compressed sensing is suggested to further mitigate HR estimation errors, particularly when frequency grid locations are mismatched.

In sum, by advancing HR estimation techniques for PPG signals amidst motion-induced noise, the research not only enhances current algorithmic approaches but also sets a foundation for future innovations in signal processing for biosignal applications.