A Stochastic Gradient Approach for Robust Heartbeat Detection With Doppler Radar Using Time-Window-Variation...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Health Monitoring

Background:

• Heart rate (HR) variability is crucial for assessing health and stress.
• Non-contact HR monitoring using Doppler radar is a growing field.
• Radar-based heartbeat detection is challenged by motion and respiration artifacts.

Purpose of the Study:

• To develop a robust non-contact heart rate estimation technique using Doppler radar.
• To improve the accuracy and stability of HR detection in the presence of noise.
• To address limitations of existing radar-based HR monitoring methods.

Main Methods:

• Proposed a zero-attracting sign least-mean-square (ZA-SLMS) algorithm for high-resolution heartbeat spectrum reconstruction.
• Introduced an adaptive regularization parameter in an improved ZA-SLMS (IZA-SLMS) for noise adaptation.
• Incorporated a time-window-variation (TWV) technique for stable HR estimation.

Main Results:

• The IZA-SLMS algorithm with TWV demonstrated significant accuracy improvements over existing methods.
• Achieved the lowest average error of 3.79 beats per minute during laptop typing.
• Successfully reconstructed accurate heartbeat spectra by correcting gradient quantization and applying sparse penalties.

Conclusions:

• The proposed IZA-SLMS algorithm with TWV offers a more stable and accurate solution for non-contact HR monitoring.
• This technique effectively mitigates performance degradation caused by respiration and body motion.
• The findings suggest a promising advancement for remote health and stress monitoring applications.