基于SVD特征降维和支持向量机的跌倒检测算法

摘要

为减少跌倒对人体造成的伤害,采用一种基于支持向量机的人体跌倒检测方法.利用安置于腰上的手机采集人体运动行为加速度数据,提取对跌倒行为敏感的时域及频域特征,利用奇异值分解方法降维特征和重构跌倒特征,采用支持向量机分类器检测跌倒行为.仿真实验表明:该方法能够有效地识别跌倒和日常行为,具有较高灵敏度和特异度,并可同时提高识别正确率.%A falling detection method based on support vector machine (SVM) is proposed to reduce the harm to human body which caused by falling down.The acceleration sensor data are collected by the mobile phone which is located on the waist, and the features which are sensitive to the falling action are extracted from the time and frequency domain.The extracted features are further processed by singular value decomposition singular value decomposition (SVD) to reduce the dimension and be rebuilt.Then, the SVM is adopted to detect the falling behavior.The simulation results show that the proposed method can effectively identify the falling behavior and the daily behavior, which achievesa higher sensitivity and specificity, improving the detection rate at the same time.