【作者】 张冲； 郭一楠； 巩敦卫；

【Author】 ZHANG Chong;GUO Yinan;GONG Dunwei;School of Information and Control Engineering,China University of Mining and Technology;Intelligent Medical Center of Artificial Intelligence Research Institute,China University of Mining and Technology;

【机构】 中国矿业大学信息与控制工程学院； 中国矿业大学人工智能研究院智慧医疗中心；

【摘要】 医疗体域网可实现人体生命体征数据的采集。然而,受体域网传感器节点的硬件故障、算力限制、供电不足或者佩戴不规范等因素的影响,体征感知数据的错误或缺失较难避免。因此,体域网故障数据检测对于提高系统可靠性十分重要。为此,本文提出一种基于约简策略与扩展置信规则库的故障检测方法。本方法采用基于马式距离的约简方法,有效地减少了训练样本;随后,根据体域网的体征属性,以最佳决策属性作为专家系统的前提属性,建立故障检测的置信框架。针对PhysioBank数据库仿真实验表明,所提算法故障检测正确率平均达97%,误报率平均仅为2%,与现有的故障算法相比,所提决策模型在大部分数据集上具有较高的准确率、召回率和F1值,并且与扩展置信规则库相比,决策模型的训练时长减小40%以上。更多还原

【Abstract】 Medical body area network can realize the collection of human vital signs data. However, due to the influence of factors such as hardware failure, computing power limitation, insufficient power supply or irregular wearing of receptor area network sensor nodes, it is difficult to avoid the error or lack of physical sign perception data. Therefore, the detection of body area network fault data is very important to improve the reliability of the system. For this reason, this paper proposes a fault detection method based on reduction strategy and extended confidence rule base. This method uses a reduction method based on horse-style distance to effectively reduce the training samples; then, according to the physical sign attributes of the body area network, the best decision attribute is used as the prerequisite attribute of the expert system to establish a confidence frame for fault detection. Simulation experiments on the PhysioBank database show that the proposed algorithm has an average fault detection accuracy rate of 97%, and an average false alarm rate of only 2%. Compared with the existing fault algorithm, the proposed decision model has a higher performance on most data sets. The accuracy rate, recall rate and F1 value of, and compared with the extended confidence rule base, the training time of the decision model is reduced by more than 40%.更多还原