基于加速度特征的人体跌倒检测算法

摘要

T he real‐time fall detection system can provide help in time for old people and pa‐tient .A fall detection algorithm based on a single triaxial accelerometer mesurement is intro‐duced .The feature of the sensor data about body′s action is analyzed ,the feature vector is con‐structed ,and the threshold value is obtained by machine leaning method .In detection process , if the feature vector of the accelerometer measurement is bigger than the threshold value ,maybe fall happened ,then the acceleration in every direction should be analyzed to determine w hether a fall is happening .T he experimental results show that the algorithm is valid .%针对跌倒检测系统必须快速准确响应的需求，提出一种基于三轴加速度传感器的人体跌倒实时检测算法。该算法通过分析人体运动的加速度特征，提取加速度信息特征向量，采用机器学习的方法对样本数据进行分析，获得分类阈值。在实时检测过程中，当采集的加速度信息的特征值大于阈值时，通过分析其产生峰值之后的加速度分量特征，确定是否发生跌倒事件。通过测试多种日常活动及各种跌倒的加速度信息，证明了该算法的有效性。