Monitoramento da saúde humana através de sensores: análise de incertezas contextuais através da teoria da evidência de Dempster-Shafer
Palavras-chave: Monitoramento da saúde humana através de sensores. Análise de Incertezas contextuais. Teoria da Evidência de Dempster-Shafer. ABSTRACT SILVA, K. C. N. Human Health Monitoring by Sensors: Analysis of Contextual Uncertainties through Dempster-Shafer Evidence Theory. 2013. 141p. Thesis (Doctoral) -Escola Politécnica da Universidade de São Paulo, 2013. The remote monitoring of human health basically involves the use of sensor network technology as a means of capturing patient data
... ring patient data and observation, in every environment. The sensor technology facilitates remote monitoring of patients with heart disease, respiratory problems, postoperative complications and even people in residential treatment. An important element of the health monitoring system is its ability to interact with the environment which allows, for example, act as a provider of relevant information and services to the user. The interaction with the environment provides to the system the characteristics related to a context-aware application, once this kind of system can react and adapt itself in face of environment´s changes, through a proactive and intelligent assistance. Another significant aspect of health monitoring systems is related to the uncertainties associated with the technology used as a means for obtaining and processing the data sensed by sensors, and the data which will be presented to the expert's users -physicians. Uncertainties are inevitable elements in any ubiquitous and context-aware application and it can be generated by incomplete or imperfect data. In the human health monitoring by sensors factors, such as the mutual influence between physiological, behavioral and environmental data are mentioned as potential generators of uncertain contextual information. This research take into consideration that each sensor captures a data type and sends it to a station located in the patient's home. The objective of this paper is to present a process to analyze the contextual uncertainties present in the monitoring of human health via sensors. The method used was based on the Dempster-Shafer Evidence Theory and The Uncertainty Factor Model. The process named PRANINC, considers each data captured, by different sensors, as evidence and, all of the evidences are considered in the formation of hypotheses. Three contextual classes of uncertainties were specified: the uncertainties arising from the technology employed in transmitting the data captured by sensors, the uncertainties related to the actual sensors, which are subject to errors and defects, and the uncertainties associated with the mutual influence between the observed variables. The method was employed through conducting experiments on files with physiological data of real patients, to which, were added behavioral and environmental factors. As a result was possible to confirm that the context influences the data transferred by the monitoring system and that contextual uncertainties may influence the quality of the information which shall be considered by the specialist.