Abstract: With the popularity of smart phones, using mobile devices to detect fall event is becoming more and more meaningful. The on-body position of mobile devices, is one kind of important context information, which affects the recognition of fall detection activity. Therefore, the paper proposes a fall detection method based on mobile device wearing position. A feature extraction method based on fusion of rotation mode component and attitude angle is proposed firstly, which uses the data sensed by the accelerometer and the gyroscope to calculate the rotation radius, the magnitude of the angular velocity as well as the attitude angle and then extracts a set of features to classify the wearing position of the mobile device by Logistic Regression (LR) model. On the basis of the wearing position, a fall detection algorithm based on time series analysis can be adjusted adaptively. Results show that the method achieves an accuracy of 95.32% on average in wearing position classification. In different positions, the accuracy of the time series analysis of the fall detection algorithm are all above 92%. Compared with the traditional fall detection method, this method has better fall detection effect in different wearing positions.

Key words: fall detection, time series analysis, mobile device, rotation mode component, attitude angle

摘要： 随着智能手机的普及，使用移动设备检测跌倒事件正变得越来越有意义。移动设备的佩戴位置作为一种重要的情境信息，影响着跌倒检测活动的识别效果。为此，提出一种移动设备佩戴位置自适应识别的人体跌倒检测方法，首先采用旋转模式分量和姿态角融合的特征提取方法，利用加速度计和陀螺仪数据计算出旋转半径、角速度幅度、姿态角并提取特征，然后用LR（Logistic Regression）模型将其分类得到移动设备的佩戴位置；随后根据位置自适应调整一种基于时序分析的跌倒检测方法。实验结果表明，该方法的移动设备佩戴位置平均识别率为95.32%，在不同位置，时序跌倒检测算法的准确率均在92%以上。与传统跌倒检测方法相比，该方法在不同佩戴位置均有更好的跌倒检测识别效果。

关键词: 跌倒检测, 时序分析, 移动设备, 旋转分量, 姿态角