Abstract: Aiming at lack of prediction of the health status currently in the monitoring system for the old people’s activities in gerocomium by low cost, by using of the data of the old people’s activities from the three-tier monitoring tracking system based on semi-active RFID which has been implemented, this article analyzes statistically and gets hold of the average number of activities which marks on health status. According to the natural relations between the number of activities and healthy status, a model of the health index and the BP neural network of three layer 8-9-1 type are designed to forecast the health index of the elderly after it has been training with the processing data. Experimental results show that the model calculation and prediction results can basically agree with the doctor’s diagnosis, false positives and non-response rates are low. The model provides a low-cost decision support for the nursing and the prediction of elderly health status, and extends effectively the RFID application for perception.

Key words: forecast, health index, elderly people, Back Propagation（BP） neural network, semi-active Radio Frequency Identification（RFID）

摘要： 针对当前敬老院老龄人的活动监控系统缺乏低成本的健康状态预测，利用半有源RFID三层监控跟踪系统获得的老龄人活动数据，统计并分析得出标示健康的平均活动次数，根据活动次数与健康的自然关系，设计了一个表征健康状态的指数模型。使用8-9-1型的三层BP网络，通过数据预处理和训练来预测老龄人的健康指数。实验结果表明该模型计算和预测结果与医生的诊断能基本相符，误报漏报率较低。该模型为老龄人的健康状态预测、健康护理提供了一种低成本的决策支持，有效扩展了RFID的感知应用。

关键词: 预测, 康指数, 老龄人, 反向传播（BP）神经网络, 半有源射频识别（RFID）