Heart Rate Imputation Using Accelerometers for Wearable Devices.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Photoplethysmography (PPG) is crucial for heart rate (HR) estimation in health monitoring.
• PPG signals are highly vulnerable to motion artifacts, leading to inaccurate HR readings and system failures.
• Existing methods struggle with reliable HR estimation during significant physical activity.

Purpose of the Study:

• To develop a novel approach for reducing catastrophic failures in PPG-based HR estimation.
• To impute missing HR data using tri-axial accelerometer data from consumer-grade smartwatches.
• To enhance the reliability of HR monitoring in real-world conditions with motion artifacts.

Main Methods:

• A lightweight neural network model was developed to leverage accelerometer data and historical HR measurements.
• The model was trained and validated on two distinct datasets containing PPG and accelerometer signals.
• The proposed method focuses on imputing missing HR values during periods of motion artifact interference.

Main Results:

• The proposed method significantly outperforms existing baseline methods in HR imputation accuracy.
• Incorporating accelerometer data during imputation proved effective in reducing HR estimation errors.
• The approach demonstrated particular efficacy during transitions in physical activity levels.

Conclusions:

• The novel HR imputation method enhances the robustness of PPG-based monitoring systems.
• Utilizing accelerometer data offers a practical solution for improving HR accuracy in smartwatches.
• This technology has the potential to make wearable health monitoring more reliable for everyday use.