Algoritmy pro detekci pádů s využitím radarových technologií

Abstract

This master's thesis deals with the design, implementation, and evaluation of algorithms for fall detection using modern radar technologies. In the theoretical part, the issue of falls is described, especially among the elderly. It also explains the fundamental principles of radar systems, their types, and possibilities for monitoring human movement. In addition, an overview of current methods and procedures for fall detection is provided, including their strengths and weaknesses. The practical part focuses on implementing and testing several detection algorithms, among which are simple threshold-based methods, fuzzy logic systems, classical machine learning algorithms, and more advanced approaches utilizing neural networks. Real and simulated datasets of measured signals from mmWave radar were used for validation, evaluating accuracy, sensitivity, the number of false alarms, and other metrics. Analysis of the results showed that modern radar technologies enable highly accurate motion differentiation and offer potential for non-invasive and reliable monitoring of individuals in both home and clinical environments. The thesis concludes by discussing the advantages and limitations of each approach and outlines possible directions for future development, including optimization for real-time deployment.