一种基于加速度传感器的人体跌倒识别方法

为了减少老年人因跌倒而造成的伤害,及时有效地识别跌倒行为,提出了一种基于三轴加速度传感器的人体跌倒识别方法。首先将加速度传感器放置于人体腰腹位置,采集人在运动时的加速度变化数据;然后使用日常活动数据训练隐马尔科夫模型(HMM),利用老年人活动状态相对较少的特点,从测量数据与HMM的匹配程度寻找"疑似"跌倒行为;最后计算短暂时间内的身体倾角,检测人体躺卧姿态,完成跌倒识别。利用HMM和身体倾角识别跌倒,解决了生活中缺乏跌倒数据训练样本的问题,提高了某些近似行为的区分度。仿真结果表明,该方法在有效识别跌倒行为的同时,提高了正确率。

Human falling detection based on accelerometer

In order to lessen the injury of elder people caused by falling, and to recognize falling timely and effectively, this paper proposed a method to identify human falling based on a tri-axis accelerometer. It put the accelerometer on the waist to collect the changing data of acceleration from human motions, trained the parameters of HMM by using the data from activities in daily life. By making use of the special characteristic of few activities of elder people, it detected suspected falling according to the matching degree between observation data and HMM. And then, it calculated body tile angle in a short time to detect human lying, which helped to complete the recognition. This method solved the problem of the inadequate training samples of falling data in daily life and improved the distinction of some similar behaviours. Simulation results show that this method can not only effectively identify human falling, but also improve the accurate rate.