TriboMotion: A Self-Powered Triboelectric Motion Sensor in Wearable Internet of Things for Human Activity Recognition and Energy Harvesting
TriboMotion: A Self-Powered Triboelectric Motion Sensor in Wearable Internet of Things for Human Activity Recognition and Energy Harvesting
复制标题
DOI：
10.1109/jiot.2018.2817841
复制
发表时间：
2018-12-01
影响因子：
10.6
通讯作者：
Sun, Ye
中科院分区：
文献类型：
--
作者：
Huang, Hui;Li, Xian;Sun, Ye
人体体力活动识别在医学诊断、福利管理、康复治疗等领域有着广泛的应用。尽管文献中提供了各种物联网(IoT)设计，但电源资源往往限制了IoT的生命周期。针对这一缺点，开发了一种新的不需要任何信号调理电路的可穿戴物联网(WIoT)中用于人体活动识别的运动传感器系统。在运动传感器的设计中，探索了基于摩擦电学的物理模型。它可以在没有任何电源的情况下采集身体活动引起的运动信号。此外，由于摩擦电结构在随机低频运动时具有较高的输出电压，在涉及连续运动时电压相对稳定，因此可以用作运动采集的能量采集器。这种新的设计为构建下一代自供电WIoT系统奠定了基础。我们的新设计经过了广泛的评估，使用了最常见的活动，包括坐和站、走路、爬上爬下和跑步。实验结果表明，该系统的识别准确率平均在80%以上，达到了与现有体力活动识别相当的性能。同时，整个传感硬件系统(包括传感器、微控制器及相应电路)的能耗降低了25%以上。
Human physical activity recognition is widely used in medical diagnosis, well-being management, and rehabilitation treatment. In spite of various Internet of Things (IoT) designs available in the literature, power resources often limit the lifetime of IoT. Regarding this weakness, this paper develops a new motion sensor system in wearable IoT (WIoT) for human physical activity recognition without any signal conditioning circuits. The triboelectricity-based physical model is explored in designing the motion sensor. It enables to collect motion signals caused by physical activities without any power supply. In addition, the triboelectric structure can be used as an energy harvester for motion harvesting due to its high output voltage in random low-frequency motion and a relatively stable voltage when involving continuous activities. Such a new design lays the foundations for constructing the next generation self-powered WIoT systems. Our new design has been extensively evaluated, where most common activities including sitting and standing, walking, climbing upstairs and downstairs, and running are used. The experimental results demonstrate that our system can achieve similar comparable performance as the state of the art for physical activity recognition at an average successful accuracy of over 80%. At the same time, our system reduces more than 25% energy consumption of the entire sensing hardware system which includes the sensor, microcontroller, and corresponding circuits.