Nmr-VSM: Non-Touch Motion-Robust Vital Sign Monitoring via UWB Radar Based on Deep Learning.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Biometric radar offers accurate, non-contact vital sign monitoring.
• Current systems are limited to static environments and single parameters.
• Collected data often remains underutilized in existing research.

Purpose of the Study:

• To develop a motion-robust, non-contact vital sign monitoring system using ultra-wideband (UWB) radar and deep learning.
• To enable multi-dimensional vital sign monitoring (heart rate, respiration rate, distance, motion status) under dynamic conditions.
• To implement cardiac anomaly detection for early disease identification.

Main Methods:

• Designed a UWB radar for multi-dimensional vital sign acquisition.
• Collected and analyzed experimental data, assessing factors like motion and distance.
• Developed a deep neural network (DNN) for motion-robust heart rate correction.
• Modeled heart rate variability (HRV) and proposed a convolutional neural network (CNN) for low-latency cardiac anomaly detection.

Main Results:

• The UWB radar system accurately monitored heart rate and respiration rate in motion environments.
• The DNN model demonstrated high robustness in correcting heart rate data during movement.
• The CNN model achieved low-latency detection of cardiac anomalies like ventricular tachycardia and fibrillation.
• Experimental validation confirmed the system's accuracy and anomaly detection capabilities.

Conclusions:

• The developed Nmr-VSM system provides accurate, multi-dimensional vital sign monitoring in motion-rich environments.
• The system enables low-latency cardiac anomaly detection, paving the way for advanced healthcare applications.
• This research addresses limitations of current non-contact monitoring systems, enhancing their practical utility.