【摘要】 以多通道融合为指导思想,综合利用多种信号处理技术针对老年人的摔倒姿态进行检测。首先介绍了九轴传感器MPU9150和脉搏血氧传感器AFE4400,然后分析了人体在不同状态下合加速度的相关特征,并由此得出人体的摔倒过程有4个不同的状态组成,分别是行走状态、自由落体状态、碰撞状态和静止状态。基于该分析,给出了用于摔倒检测的方案。先通过加速计,初步判断是否有摔倒发生,在此基础上,利用了陀螺仪和磁力计,判断人体姿态是否出现了明显的改变,由此辅助判断摔倒,降低虚警率。最后,由于老年人在摔倒时,往往会伴随短时间的脉搏的突变,给出了利用脉搏进一步的摔倒检测,从而可大大提升摔倒检测的准确率。更多还原

【Abstract】 A method for human fall detection based on multichannel mixture is proposed. Firstly,two sensors,MPU9150 and AFE4400,are introduced which are used in this scheme. Secondly,the characteristics of the acceleration of the human body in different states are analyzed. Based on this analysis,fall procedure can be divided into four stages: normal walk state,free fall state,collusion state and zero motion state. Using this conclusion,the fall detection scheme consists of three steps: first step is to use accelerometer to determine whether the fall occurred; second step is to use gyroscope and magnetometer to determine whether the human posture has changed significantly,this information can be used to reduce false alarm rate of fall detection. In the last step,since the pulse should be increased dramatically after fall is happened,this paper uses AFE4400 to detect the change of pulse; this step also can reduce the false alarm rate and ensure feasibility of the whole scheme.更多还原