A Movement Detection System Using Continuous-Wave Doppler Radar Sensor and Convolutional Neural Network to Detect...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Sensor Technology

Background:

• The COVID-19 pandemic highlights the need for effective, non-contact monitoring of respiratory symptoms.
• Artificial intelligence, particularly convolutional neural networks (CNNs), offers advanced capabilities for analyzing sensor data.
• Traditional imaging methods raise privacy concerns and are limited by environmental conditions.

Purpose of the Study:

• To develop and evaluate a novel system for detecting coughs, an important COVID-19 symptom, using a K-band radar sensor and CNNs.
• To assess the accuracy of different CNN architectures (AlexNet, VGG-19, GoogLeNet) for cough detection at varying distances.
• To demonstrate the feasibility of a camera-free, privacy-preserving, and contactless cough detection system.

Main Methods:

• Utilized a K-band continuous-wave Doppler radar sensor to capture cough sounds.
• Applied popular convolutional neural network (CNN) architectures including AlexNet, VGG-19, and GoogLeNet for cough detection.
• Evaluated system performance at distances of 1m and 3m, as well as with a mixed dataset.

Main Results:

• Achieved a cough detection test accuracy of 88.0% with AlexNet at 1m distance.
• Demonstrated 80.0% accuracy with AlexNet at 3m distance and 86.5% with a mixed dataset.
• The radar sensor proved to be inexpensive, camera-free, and independent of lighting conditions.

Conclusions:

• The proposed radar-based CNN system offers a robust, privacy-preserving method for contactless cough detection.
• This technology is suitable for monitoring contagious diseases like COVID-19 in various environments.
• The system's environmental robustness and independence from lighting conditions make it superior to traditional cameras.