Kim Duk-Jin, Suk Myoung Hoon, Prabhakaran B
University of Texas at Dallas, Richardson, TX 80305, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2012;2012:5234-7. doi: 10.1109/EMBC.2012.6347174.
Pervasive computing becomes very active research field these days. A watch that can trace human movement to record motion boundary as well as to study of finding social life pattern by one's localized visiting area. Pervasive computing also helps patient monitoring. A daily monitoring system helps longitudinal study of patient monitoring such as Alzheimer's and Parkinson's or obesity monitoring. Due to the nature of monitoring sensor (on-body wireless sensor), however, signal noise or faulty sensors errors can be present at any time. Many research works have addressed these problems any with a large amount of sensor deployment. In this paper, we present the faulty sensor detection and isolation using only two on-body sensors. We have been investigating three different types of sensor errors: the SHORT error, the CONSTANT error, and the NOISY SENSOR error (see more details on section V). Our experimental results show that the success rate of isolating faulty signals are an average of over 91.5% on fault type 1, over 92% on fault type 2, and over 99% on fault type 3 with the fault prior of 30% sensor errors.
如今，普适计算成为了一个非常活跃的研究领域。一种能够追踪人体运动以记录运动边界，并通过个人的局部访问区域来研究社交生活模式的手表。普适计算也有助于患者监测。一个日常监测系统有助于对诸如阿尔茨海默病、帕金森病或肥胖症监测等患者监测进行纵向研究。然而，由于监测传感器（可穿戴无线传感器）的特性，信号噪声或传感器故障误差可能随时出现。许多研究工作通过大量的传感器部署来解决这些问题。在本文中，我们提出仅使用两个可穿戴传感器进行故障传感器检测与隔离。我们一直在研究三种不同类型的传感器误差：SHORT误差、CONSTANT误差和NOISY SENSOR误差（详见第五节）。我们的实验结果表明，在传感器误差先验为30%的情况下，对于故障类型1，隔离故障信号的成功率平均超过91.5%；对于故障类型2，超过92%；对于故障类型3，超过99%。
