Background and objective Chronic cardiovascular diseases have motivated the development of wearable systems capable of continuously monitoring physiological variables such as oxygen saturation (SpO<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$_2$</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mn>2</mn> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation>) and non-invasive blood pressure (NIBP). However, motion artifacts remain one of the main limitations affecting the reliability of these measurements in ambulatory environments. This article presents a novel framework for motion artifact reduction in photoplethysmography (PPG) and NIBP signals based on redundant sensing, inertial measurements, ECG-guided temporal synchrony, and multichannel signal processing. Methods The proposed framework combines independent component analysis and recursive least-squares adaptive filtering to separate physiological information from motion-induced interference. Performance was compared with conventional finite impulse response (FIR) filtering and wavelet shrinkage (WS) methods using signal-to-noise ratio (SNR), weighted distortion assessment (WDA), oxygen saturation estimation, and blood pressure estimation metrics. Results The proposed method achieved statistically significant improvements in signal quality and physiological parameter estimation. Average SNR improved from <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$-5.93\pm 5.05$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo>-</mo> <mn>5.93</mn> <mo>±</mo> <mn>5.05</mn> </mrow> </math></EquationSource> </InlineEquation> dB to <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$-3.43\pm 2.02$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo>-</mo> <mn>3.43</mn> <mo>±</mo> <mn>2.02</mn> </mrow> </math></EquationSource> </InlineEquation> dB in the red PPG channel (<InlineEquation ID="IEq4"> <EquationSource Format="TEX">$p=0.02$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>p</mi> <mo>=</mo> <mn>0.02</mn> </mrow> </math></EquationSource> </InlineEquation>), while the WDA index increased from 0.48 to 0.73. Oxygen saturation estimation improved from <InlineEquation ID="IEq5"> <EquationSource Format="TEX">$80.87\pm 9.09\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>80.87</mn> <mo>±</mo> <mn>9.09</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> in motion-contaminated signals to <InlineEquation ID="IEq6"> <EquationSource Format="TEX">$94.16\pm 2.72\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>94.16</mn> <mo>±</mo> <mn>2.72</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> after denoising (<InlineEquation ID="IEq7"> <EquationSource Format="TEX">$p=0.04$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>p</mi> <mo>=</mo> <mn>0.04</mn> </mrow> </math></EquationSource> </InlineEquation>), approaching resting measurements (<InlineEquation ID="IEq8"> <EquationSource Format="TEX">$95.23\pm 2.31\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>95.23</mn> <mo>±</mo> <mn>2.31</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation>). For NIBP signals, statistically significant improvements were also observed (<InlineEquation ID="IEq9"> <EquationSource Format="TEX">$p=0.045$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>p</mi> <mo>=</mo> <mn>0.045</mn> </mrow> </math></EquationSource> </InlineEquation>), particularly for systolic pressure estimation, although diastolic pressure remained affected by residual motion-related distortions. Conclusions The proposed framework improves the robustness of wearable PPG and NIBP monitoring under motion conditions through the combined use of sensor redundancy, ECG-guided temporal coupling, inertial measurements, and adaptive multichannel signal processing. The results demonstrate significant improvements in signal quality and physiological parameter estimation compared with conventional denoising approaches, supporting the potential use of the method in ambulatory cardiovascular monitoring applications.

Reduction of Motion Artifacts in Cardiorespiratory Vital Signs as Oxygen Saturation and Non-Invasive Blood Pressure Through Redundant Denoising and Adaptive Filtering Methods for Wearable Healthcare Monitoring Systems

期刊论文
出版模式：

开放获取
发表日期：

2026年8月3日

查看全文

Fabian Andres Castaño,
Alher Mauricio Hernández

摘要

Background and objective

Chronic cardiovascular diseases have motivated the development of wearable systems capable of continuously monitoring physiological variables such as oxygen saturation (SpO $_2$ $_{2}$ ) and non-invasive blood pressure (NIBP). However, motion artifacts remain one of the main limitations affecting the reliability of these measurements in ambulatory environments. This article presents a novel framework for motion artifact reduction in photoplethysmography (PPG) and NIBP signals based on redundant sensing, inertial measurements, ECG-guided temporal synchrony, and multichannel signal processing.

Methods

The proposed framework combines independent component analysis and recursive least-squares adaptive filtering to separate physiological information from motion-induced interference. Performance was compared with conventional finite impulse response (FIR) filtering and wavelet shrinkage (WS) methods using signal-to-noise ratio (SNR), weighted distortion assessment (WDA), oxygen saturation estimation, and blood pressure estimation metrics.

Results

The proposed method achieved statistically significant improvements in signal quality and physiological parameter estimation. Average SNR improved from $-5.93\pm 5.05$ $- 5.93 \pm 5.05$ dB to $-3.43\pm 2.02$ $- 3.43 \pm 2.02$ dB in the red PPG channel ( $p=0.02$ $p = 0.02$ ), while the WDA index increased from 0.48 to 0.73. Oxygen saturation estimation improved from $80.87\pm 9.09\%$ $80.87 \pm 9.09 %$ in motion-contaminated signals to $94.16\pm 2.72\%$ $94.16 \pm 2.72 %$ after denoising ( $p=0.04$ $p = 0.04$ ), approaching resting measurements ( $95.23\pm 2.31\%$ $95.23 \pm 2.31 %$ ). For NIBP signals, statistically significant improvements were also observed ( $p=0.045$ $p = 0.045$ ), particularly for systolic pressure estimation, although diastolic pressure remained affected by residual motion-related distortions.

Conclusions

The proposed framework improves the robustness of wearable PPG and NIBP monitoring under motion conditions through the combined use of sensor redundancy, ECG-guided temporal coupling, inertial measurements, and adaptive multichannel signal processing. The results demonstrate significant improvements in signal quality and physiological parameter estimation compared with conventional denoising approaches, supporting the potential use of the method in ambulatory cardiovascular monitoring applications.