基于卡尔曼滤波与k-NN算法的可穿戴跌倒检测技术研究

发布时间：2018-04-14 16:17

本文选题：计算机应用技术 + 跌倒检测　；参考：《电子与信息学报》2017年11期

【摘要】：针对老年人跌倒检测的准确性和实时性需求,该文首先建立了基于姿态角的活动描述模型,研发了集成加速度传感器、陀螺仪和蓝牙的活动感知模块,从而实时采集运动变化数据并使用蓝牙发送到智能手机。其次,选取姿态角及加速度信号向量模作为特征量,通过卡尔曼滤波对数据进行去噪与融合,并应用滑动窗口和k-NN算法实现了可实时感知老年人跌倒并报警的系统。实验证明系统在二分类场景下的跌倒检测准确率为98.9%,而敏感度和特异性分别达到98.9%和98.5%,验证了系统具有良好的实时性和较高的准确率。
[Abstract]:The fall detection accuracy and real-time demand for the elderly, this paper established a description model of the attitude angle based on the research of integrated accelerometer, gyroscope and active sensing module of Bluetooth, and real-time data of the movement and the use of Bluetooth mobile phone to send intelligent. Secondly, selection of attitude angle and acceleration signal as a vector module based on the data features, Calman filtering denoising and fusion, and the application of sliding window and the k-NN algorithm of the system can be real-time perception of falls in elderly people and the alarm. The experiment proved that the accuracy of fall detection in the two scenarios of the system is 98.9%, and the sensitivity and specificity were respectively 98.9% and 98.5%, verified the accuracy of system has good real-time and high.

【作者单位】：北京未来网络科技高精尖创新中心;北京市物联网软件与系统工程技术研究中心;
【基金】：国家自然科学基金(61602016)~~
【分类号】：TP212.9;TP274

【相似文献】