Abstract:

The health smart home（HSH） is an important measure to ease up the social pressure caused by the aging population and to optimize the allocation of medicine recourses.In an HSH,the vital parameters of a patient can be monitored and automatically analyzed at home.This paper presents an analytical method of monitoring data based on the time series modeling.Three modules of the proposed method,namely the model identification,the model adjustment and the prediction interval（PI） determination are discussed.In the proposed method,the model order is determined according to the final prediction error（FPE） criterion,thus ensuring the model to accord well with the monitoring data.The model parameters are adjusted on line,based on adaptive filter algorithms,thus facilitating the model to describe the dynamic features of monitoring data in a better way.The interval of 30-step-forward prediction is computed according to the modeling results,thus implementing the recognition of the characteristic patterns with stable data,outlier data and state change,respectively.Moreover,three datasets in PhysioNet biomedicine database are used to perform an experimental investigation.The results indicate that the proposed method can analyze the continuous monitoring data on line with high accuracy.

Key words: health smart home, monitoring, time series modeling, final prediction error criterion, adaptive filter