A Contactless and Fine-Grained Sleep Monitoring System Leveraging WiFi Channel Response
A Contactless and Fine-Grained Sleep Monitoring System Leveraging WiFi Channel Response
复制标题
DOI：
10.1109/icc.2019.8761602
复制
发表时间：
2019-05
期刊：
影响因子：
--
通讯作者：
Y. Gu;Chenyu Zhang;Yantong Wang;Zhi Liu;Yusheng Ji;Jie Li
中科院分区：
文献类型：
--
作者：
Y. Gu;Chenyu Zhang;Yantong Wang;Zhi Liu;Yusheng Ji;Jie Li
如何在没有专门硬件的情况下，有效记录由静止姿势和原地运动组成的细粒度睡眠记录，用于睡眠障碍诊断？现有的基于传感器或基于视觉的解决方案要么妨碍使用，要么依赖特定设备。本文介绍了 SleepGuardian，这是一种基于射频 (RF) 的睡眠监测系统，仅利用无处不在的 WiFi 信号来提供安静（不引人注目且无隐私问题）但忠诚（细粒度且可靠）的记录服务。 SleepGuardian 的关键是将无线信道的能量特征建模为高斯混合模型 (GMM)，以自适应地识别睡眠期间发生的运动。我们使用现成的 WiFi 设备对 SleepGuardian 进行原型设计，并在办公室中对其进行评估。对 11 名受试者进行的几次人工和真实睡眠周期的实验结果表明，SleepGuardian 是有效的，因为它平均达到 100% 的总体准确率 (ACC)、0% 的假阴性率 (FNR) 和 0.64 秒的平均绝对误差 (MAE)。考虑到 SleepGuardian 与现有 WiFi 基础设施兼容，它构成了一种低成本但前景广阔的睡眠监测解决方案。
How can we effectively log a fine-grained sleep record consisting of still postures and in-place motions for the sleep disorder diagnosis without any specialized hardware? Existing sensor-based or vision-based solutions are either obstructive to use or rely on particular devices. This paper introduces SleepGuardian, a Radio Frequency (RF) based sleep monitoring system leveraging only omnipresent WiFi signals to provide a silent (unobtrusive and free of privacy concerns) yet loyal (finegrained and reliable) logging service. The key to SleepGuardian is to model the energy feature of wireless channel as a Gaussian Mixture Model (GMM) to adaptively recognize motions happened during sleep. We prototype SleepGuardian with off-the-shelf WiFi devices and evaluate it in an office. Experimental results over 11 subjects with several artificial and real periods of sleep demonstrate that SleepGuardian is effective since it achieves 100% overall accuracy (ACC), 0% false negative rate (FNR) and 0.64 s mean absolute error (MAE) on average. Considering that SleepGuardian is compatible with existing WiFi infrastructure, it constitutes a low-cost yet promising solution for sleep monitoring.