智能健康住宅监护数据分析方法的研究

智能健康住宅为缓解人口老龄化的压力和优化医疗资源配置提供了重要手段。其特点是在家庭环境下提供各项生命体征参数监测，并对监护数据自动分析处理。本文提出了以时序建模为基础的分析方法。此方法由三个部分构成，包括模型辨识、模型更新、以及基于模型的预报区间确定。模型阶数基于最终预报准则确定，使得模型能更好的符合观测数据。基于自适应滤波器算法对模型参数进行在线更新，确保模型能更好的描述监护数据的动态特性。根据建模结果，作前向30步预报，确定预报区间。从而实现对监护数据中的平稳、异常值、以及状态变化三种特征模式的识别。通过应用PhysioNet中的三组数据集进行实验研究。实验结果表明此方法的预报结果准确，能实现对于智能健康住宅中连续监测获得的数据进行在线分析。

Study on Analysis Methods for Monitoring Data in Health Smart Home

Health Smart Home (HSH) provided important solutions for aging people and optimal allocation of medicine recourses. HSH characteristics were that monitored patient’s vital parameters at home and analyze the monitoring data. The methods for monitoring data analysis were proposed in this paper. These methods based on time series modeling, and included model identification, model adjustment, and boundaries of the prediction intervals (PI) computation. The model’s order was determined based on the final prediction error (FPE) criterion, and then kept the model agreeing sufficiently well with the observed data. The model’s parameters were adjusted on-line based on adaptive filter algorithms, and then kept the model describing the true system of time series monitoring vital signs data. The PI of 30-steps-forward prediction was computed for three characteristic patterns’ detection, which was no change, outlier, and level change. Datasets from PhysioNet biomedicine database were used for test. The results prove that these methods can realize correct prediction and can be used for vital signals data processing on-line in HSH.