Toward Accurate Respiratory Rate Estimation Using Hybrid Adaptive Filters on Smartwatch PPG

David Pollreisz, Mostafa Haghi, Nima TaheriNejad

Confirmation: I have read and agree with the IEEE BHI 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: photoplethysmography signal, respiratory rate estimation, adaptive filtering, synthetic noise

Abstract:

The growing adoption of wearable devices has opened new opportunities for unobtrusive health monitoring
and medical research. Among various vital signs, Respiratory Rate (RR), often referred to as the forgotten vital sign, is
a key parameter, and estimating it using smartwatch-based Photoplethysmogram (PPG) offers a convenient and nonintrusive alternative to conventional techniques. However, motion artifacts significantly impair signal quality, posing a major challenge for accurate RR estimation. In this study, we present a novel algorithm that combines adaptive filtering by employing a Least Mean Squares (LMS) filter guided by a reference signal derived from Empirical Mode Decomposition (EMD), supported
by frequency-domain peak detection to robustly estimate RR. Experimental results demonstrate a 71% reduction in mean
estimation error and a 56% reduction in standard deviation compared to both adaptive and non-adaptive baseline methods.
These improvements show the enhanced accuracy and reliability of the proposed approach. The low estimation error makes this method a practical solution for continuous RR monitoring in everyday settings, enabling broader integration into wearable healthcare applications.

Track: 7. General Track

Registration Id: TXNDQSSZY5S

Submission Number: 295