Radio frequency (RF) sensing for deep awareness of human physical status

• Main Author: Fang, Chenhao
• Other Authors: Luo Jun
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Deep learning; Breathing pattern recognition; Data preprocessing; Radio frequency(RF) sensing
• Online Access: https://hdl.handle.net/10356/174993
• Institution: Nanyang Technological University

This project presents an innovative exploration into the application of Radio Frequency (RF) Sensing for monitoring human physical status in a non-invasive manner. The primary focus was on the development of a system capable of accurately detecting and analysing human breathing pattern used advanced and costless RF sensing technology. This was achieved by integrated RF sensing with cutting-edge digital signal processing techniques and deep learning models to create a robust and sensitive monitoring platform. The studied begins by establishing a foundational understanding of RF sensing principles, followed by a detailed examination of various RF sensors, included the NeuLog Respiration Monitor Belt and the Novelda Xethru X4 UWB Radar. These sensors play a pivotal role in capturing precise physiological data. The project then delves into the implementation of deep learning algorithms, which were crucial for processing and interpreting the complex data gathered by the RF sensors. In particular, the performance of a 1D CNN model was evaluated over 10 runs, achieved an average validation loss of 0.1526 and a validation accuracy of 95.06%, demonstrating the model's efficacy in signal analysis. The system's effectiveness was demonstrated through its ability to distinguish between different types of breathing patterns, showcasing potential applications in real-time health monitoring and acute illness detection. The research highlighted the advantages of using non-invasive methods in healthcare, emphasizing the system's suitability for continuous monitoring without causing discomfort or requiring active patient cooperation. Challenges encountered during the project, such as data collection constraints,computational limitations, and the need for extensive training data for the deep learning models, are discussed. The project concludes by outlining future directions, including refining the system for broader applications in various healthcare settings, enhancing the model's accuracy with larger datasets, and exploring the integration of this technology in other domains like infant care and IoT-based health systems. This project not only contributes to the field of healthcare technology by providing a viable solution for non-invasive monitoring but also opens avenues for further research and development in the integration of RF sensing and deep learning for health diagnostics.