mmWave-RM: A Respiration Monitoring and Pattern Classification System Based on mmWave Radar.

Area of Science:

• Biomedical Engineering
• Medical Devices
• Signal Processing

Background:

• Abnormal breathing patterns can signify serious cardiopulmonary conditions.
• Early detection of respiratory abnormalities is crucial for preventing disease progression.
• Non-contact monitoring methods are needed for continuous and unobtrusive patient assessment.

Purpose of the Study:

• To develop and validate a non-contact respiratory monitoring system using millimetre-wave (mmWave) radar.
• To accurately classify different breathing patterns, including normal, slow/deep, quick, and meningitic breathing.
• To mitigate noise interference for reliable respiratory signal detection in daily environments.

Main Methods:

• Utilized a 77 GHz frequency-modulated continuous wave (FMCW) mmWave radar for non-contact signal acquisition.
• Applied signal superposition and elliptical filtering (0.1-0.5 Hz) to reduce static noise.
• Employed Histogram of Oriented Gradient (HOG) feature extraction with K-nearest neighbours (KNN), Convolutional Neural Network (CNN), and G-SVM for breathing mode classification.

Main Results:

• Achieved high accuracy in classifying four distinct breathing modes.
• Demonstrated the effectiveness of mmWave radar in capturing subtle respiratory signals.
• Successfully filtered environmental noise to enhance signal clarity and waveform accuracy.

Conclusions:

• The mmWave radar system provides an effective solution for non-contact respiratory monitoring.
• The developed classification algorithms accurately identify various breathing patterns.
• This technology supports enhanced daily medical monitoring and early disease detection.