Radar sensing for ambient assisted living application with artificial intelligence

Li, Zhenghui (2024) Radar sensing for ambient assisted living application with artificial intelligence. PhD thesis, University of Glasgow.

Abstract

In a time characterized by rapid technological advancements and a noticeable trend towards an older average population, the need for automated systems to monitor movements and actions has become increasingly important. This thesis delves into the application of radar, specifically Frequency Modulated Continuous Wave (FMCW) radar, as an emerging and effective sensor in the field of "Activity Recognition." This area involves capturing motion data through sensors and integrating it with machine learning algorithms to autonomously classify human activities. Radar is distinguished by its ability to accurately track complex bodily movements while ensuring privacy compliance.

The research provides an in-depth examination of FMCW radar, detailing its operational principles and exploring radar information domains such as range-time and micro-Doppler signatures. Following this, the thesis presents a state-of-the-art review in activity recognition, discussing key papers and significant works that have shaped the field. The thesis then focuses on research topics where contributions were made. The first topic is human activity recognition (HAR) with different physiology, presenting a comprehensive experimental setup with radar sensors to capture various human activities. The analysis of classification results reveals the effectiveness of different radar representations. Advancing into the domain of resource-constrained system platforms. It introduces adaptive thresholding for efficient data processing and discusses the optimization of these methods using artificial intelligence, particularly focusing on the evolution algorithm such as Self-Adaptive Differential Evolution Algorithm (SADEA). The final chapter discusses the use of Long Short-Term Memory (LSTM) networks for short-range personnel recognition using radar signals. It details the training and testing methodologies and provides an analysis of LSTM networks performance in temporal classification tasks.

Overall, this thesis demonstrates the effectiveness of merging radar technology with machine learning in HAR, particularly in assisted living. It contributes to the field by introducing methods optimized for resource-limited settings and innovative approaches in temporal classification using LSTM networks.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering
• Supervisor's Name: Le Kernec, Dr. Julien and Zhang, Professor Lei
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84245
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 17 Apr 2024 15:48
• Last Modified: 17 Apr 2024 15:49
• Thesis DOI: 10.5525/gla.thesis.84245
• URI: https://theses.gla.ac.uk/id/eprint/84245

Related URLs

• Enlighten Publications Record
• Article DOI
• Article DOI
• Article DOI
• Conference proceeding
• Conference proceeding
• Conference proceeding