Abstract: In order to reduce the harm caused by fall in the elderly and detect the fall in real time, a fall detection system based on Android smartphone is designed and developed. The proposed algorithm combines the fall detection with gesture recognition algorithm for identifying daily activities and fall. The alarming information will be sent with the user’s position obtained from  GPS when falling is detected. The results of simulation and experiments show that the system can effectively distinguish between falls and daily behaviour, and the algorithm has high instantaneity, sensitivity and specificity.

Key words: fall detection, smartphone, acceleration sensor, posture recognition

摘要： 为减少跌倒对老年人造成的伤害，并对跌倒进行实时检测，提出了一种基于Android智能手机的人体跌倒检测系统，手机安置于腰上采集手机加速度传感器数据，利用了姿态识别和跌倒检测相结合的算法，区分出跌倒行为和人体日正常常活动。当检测到异常跌倒时，报警信息以及从手机中GPS获取的位置被发送。仿真及实验表明：系统能够有效地识别出跌倒和日常行为，算法具有较高实时性、具有较高灵敏度和特异度。

关键词: 跌倒检测, 智能手机, 加速度传感器, 姿态识别