基于可穿戴设备的跌倒检测算法综述

综述2013-2018年基于穿戴式惯性传感器的跌倒检测研究工作.从跌倒的定义出发，阐述常规跌倒行为的几种状态、跌倒的分类方式及其类别.以可穿戴跌倒检测系统框架为基础，依次从数据采集、预处理、特征提取、模型构建等角度分别展开介绍当前的研究工作.归纳用于跌倒检测性能评估的一系列技术指标，展示9个跌倒检测的公开数据集，以及当前跌倒检测研究工作在这些数据集上的检测精度.旨在为后续开展可穿戴跌倒检测工作提供借鉴与参考.
Abstract: Fall detection methods based on wearable inertial devices were elaborated from 2013 to 2018. First of all, the definition of fall, conventional phases contained within a fall, classification and categories of falls were fully introduced. Secondly, current research works were introduced with respect to modules such as data collection, preprocessing, feature extraction and model construction of the wearable fall detection system framework. A series of widely-used technical criteria were induced for evaluating the performance of fall detection methods. At last, nine public fall detection datasets were described, as well as the predictive performance based on those datasets, which is helpful for future research in fall detection research area.