Research on Ultra-Wideband Radar Echo Signal Processing Method Based on P-Order Extraction and VMD

Qingjie Qi, Youxin Zhao, Liang Zhang, Zhen Yang, Lifeng Sun, Xinlei Jia2022 · DOI

TL;DR

This paper introduces a novel method for non-contact vital sign monitoring using Ultra-Wideband (UWB) radar. By combining a new "P-order extraction" technique with Variational Mode Decomposition (VMD), the researchers significantly improve the accuracy of detecting respiration and heartbeat signals, even in complex environments with interference.

What Problem Does This Paper Address?

Detecting vital signs like breathing and heartbeat from a distance, especially in challenging scenarios such as post-disaster rescue or through walls, is a critical area of research. While UWB radar offers advantages like non-contact monitoring and penetration capabilities, its echo signals are often plagued by significant interference and noise. Existing methods struggle to accurately extract subtle vital sign information from these noisy signals, leading to reduced accuracy and robustness. This paper aims to overcome these limitations by developing a more effective signal processing approach for UWB radar vital sign detection.

Methods

The proposed method consists of two main stages:

  • P-order Extraction: This novel technique is designed to extract strong physical sign information from the UWB radar echo signals. It involves a multi-step process that identifies and isolates key signal components related to respiration and heartbeat. The paper details a specific implementation where "P=8" was found to be an optimal parameter for this extraction, balancing accuracy and processing time. This method aims to pre-process the signal to enhance the vital sign components before further decomposition.
  • Variational Mode Decomposition (VMD): After the P-order extraction, VMD is employed to decompose the enhanced signal into several intrinsic mode functions (IMFs). VMD is a powerful signal decomposition technique that adaptively separates a signal into its constituent frequency components. The key advantage of VMD here is its ability to handle non-linear and non-stationary signals, which are characteristic of vital signs. By decomposing the signal into IMFs, the researchers can then isolate and reconstruct the respiration and heartbeat signals based on their characteristic frequency bands.

The paper also outlines several echo signal preprocessing steps to further clean the data before the main processing:

  • Channel Signal Subtraction
  • Time Average Subtraction
  • Linear Trend Suppression
  • Signal Enhancement (using autocorrelation)

Datasets

The paper does not explicitly mention using pre-existing public datasets for training or evaluation. Instead, it describes:

  • Experimental Platform Data: Data was collected using a custom-built UWB radar system and an experimental platform that simulated human breathing using a micro-vacuum pump and a balloon. This allowed for controlled testing of distance and frequency measurements.
  • Actual Scene Data: The algorithm was also tested on a volunteer in a controlled experimental scene to validate its performance in a more realistic setting.

Key Insights and Ideas

  • Synergy of P-order Extraction and VMD: The core insight is that combining a specialized signal extraction method (P-order) with a robust decomposition technique (VMD) leads to superior vital sign detection compared to using either method alone or traditional approaches.
  • Optimizing P-order Extraction: The paper highlights the importance of parameter tuning for the P-order extraction, with P=8 identified as a sweet spot for efficient and accurate vital sign extraction. This suggests that careful selection of parameters can significantly impact performance.
  • VMD's Adaptability: VMD's ability to decompose signals into adaptive IMFs is crucial for separating the distinct frequency components of respiration and heartbeat, even when they are embedded in noisy radar echoes.
  • Frequency Band Selection for Reconstruction: The method relies on identifying IMFs whose center frequencies fall within the known ranges of human respiration (0.2-0.5 Hz) and heartbeat (0.8-2.5 Hz) for accurate signal reconstruction.

Key Results

  • Improved Accuracy: The proposed method effectively eliminated respiratory clutter and achieved accurate measurements of respiratory and heartbeat signals.
  • Experimental Platform Results:
    • Average relative error for distance measurement: 1.17%.
  • Average relative error for simulated respiratory frequency: 4.47%.
  • Actual Scene Results (Volunteer Test):
    • Average relative error for distance measurement: 1.52%.
  • Average relative error for respiratory frequency: 2.92%.
  • Average relative error for heartbeat frequency: 4.24%.
  • Parameter Optimization: The study found P=8 to be a suitable parameter for P sub-strong sign extraction, leading to higher accuracy in breathing and heartbeat detection.

Limitations and Future Work

  • Limitations:
    • The accuracy of the simulated lung in the experimental platform was noted as a source of error.
  • The paper acknowledges that the heartbeat signal is more difficult to extract than the respiratory signal, leading to slightly higher error rates for heartbeat detection.
  • The performance can be affected by the distance to the target, with errors increasing at greater distances.
  • The authors mention that the volunteer's slight movements during measurement could also introduce errors.
  • Future Work:
    • The authors plan to further optimize the details of the P-order extraction process.
  • They aim to introduce more adjustable parameters to broaden the algorithm's application scenarios.
  • Further optimization of the main parameters for VMD decomposition and more reasonable parameter selection are also planned.
  • The goal is to quantify the decomposition results more precisely.

Summary updated

BibTeX
@misc{qi2022research,
  author = {Qingjie Qi and Youxin Zhao and Liang Zhang and Zhen Yang and Lifeng Sun and Xinlei Jia},
  title = {{Research on Ultra-Wideband Radar Echo Signal Processing Method Based on P-Order Extraction and VMD}},
  year = {2022},
  doi = {10.3390/s22186726},
}
Figure