面向老年人的跌倒检测方法研究进展及展望

A Review and Prospect of Fall Detection Methods for the Elderly

摘要: 跌倒检测是老年人健康监测和安全保障领域的研究热点。随着全球人口老龄化趋势的加剧，65岁以上老年人中约有三分之一每年至少跌倒一次，跌倒后若不能及时获得救助可能导致严重后果甚至死亡，因此开发高效、准确的跌倒检测系统对于保障老年人安全具有重要意义。近年来，随着计算机视觉、深度学习、传感器技术和雷达技术的快速发展，跌倒检测技术取得了显著进展，从传统的基于阈值和规则的方法发展到基于深度学习的智能方法，从接触式传感器发展到非接触式视觉和雷达系统，呈现出多样化的发展趋势。本文基于近年来发表的相关研究论文，对跌倒检测领域的研究进展进行系统综述，从基于深度学习的视觉检测方法(重点综述基于YOLO系列目标检测算法的跌倒检测方法)、基于雷达的非接触检测方法、基于传感器的检测方法以及多模态融合检测方法等多个角度对现有研究进行分类和总结，详细分析了各类方法的技术原理、实现方案、性能特点和适用场景，比较了不同方法的优缺点，并对跌倒检测领域面临的数据隐私保护、复杂环境适应性、误报率控制等挑战和未来发展趋势进行了展望，为该领域的进一步研究提供参考。

Abstract: Fall detection is a research hotspot in the field of health monitoring and safety assurance for the elderly. With the intensification of global population aging, approximately one-third of people aged 65 and above fall at least once a year. Failure to obtain timely assistance after a fall may lead to serious consequences or even death. Therefore, the development of efficient and accurate fall detection systems is of great significance for ensuring the safety of the elderly. In recent years, with the rapid development of computer vision, deep learning, sensor technology and radar technology, remarkable progress has been made in fall detection technology. It has evolved from traditional threshold- and rule-based methods to intelligent methods based on deep learning, and from contact sensors to non-contact vision and radar systems, showing a diversified development trend. Based on relevant research papers published in recent years, this paper systematically reviews the research progress in the field of fall detection. It classifies and summarizes existing studies from multiple perspectives, including deep learning-based visual detection methods (focusing on fall detection methods based on the YOLO series of object detection algorithms), radar-based non-contact detection methods, sensor-based detection methods, and multi-modal fusion detection methods. It also analyzes in detail the technical principles, implementation schemes, performance characteristics and applicable scenarios of each type of method, compares the advantages and disadvantages of different methods, and discusses the challenges faced by the field of fall detection, such as data privacy protection, complex environment adaptability, and false alarm rate control, as well as future development trends, providing a reference for further research in this field.