Time-variant traits analysis in respiratory doppler radar’s signal
Doppler radar-based respiratory monitoring offers a non-contact, physiologic assessment of breathing patterns. However, the inherent time-variant nature of respiratory signals presents challenges in accurate characterisation and classification. This study investigates the analysis of time-variant tr...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Polska Akademia Nauk
2026
|
| Online Access: | http://eprints.utem.edu.my/id/eprint/29604/2/0264326022026939333054.pdf http://eprints.utem.edu.my/id/eprint/29604/ https://ijet.pl/index.php/ijet/article/view/10.24425-ijet.2026.157901 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Doppler radar-based respiratory monitoring offers a non-contact, physiologic assessment of breathing patterns. However, the inherent time-variant nature of respiratory signals presents challenges in accurate characterisation and classification. This study investigates the analysis of time-variant traits in respiratory Doppler radar signals using a feature extraction framework that integrates statistical features, Hilbert transform, discrete wavelet transforms (DWT), and fractal dimension analysis. The methodology begins with signal pre-processing to remove noise and enhance the signal for clarity. Statistical features, including mean, skewness, and kurtosis, are extracted to quantify signal variability. The Hilbert transform is employed to analyse instantaneous amplitude and phase variations, while DWT is used for multi-resolution decomposition to capture respiratory signal dynamics across different frequency scales over time. Additionally, fractal dimension analysis provides insights into the complexity and irregularity of breathing patterns in the time series. Machine learning-based classification models are applied to distinguish between normal and abnormal respiratory conditions. Results demonstrate the effectiveness of the proposed approach in enhancing respiratory signal characterisation and classification by utilising the Hilbert Transform over a Subspace Discriminant model with an accuracy rate of 92.3%. The findings suggest that integrating these feature extraction techniques can significantly improve non-invasive respiratory monitoring. |
|---|