Context Reasoning Under Uncertainty Based On Evidential Fusion Networks In Home-based Care
MetadataShow full item record
Pervasive computing technologies use embedded intelligent systems to enable various real-time applications. Some of these applications are: continuous healthcare monitoring, autonomous diagnosis and treatment, and remote disease management without spatial-temporal limitations. Additional healthcare applications include home-based care, disaster relief management, medical facility management, and sports health management. Issues related to the pervasive healthcare are generally classified into five categories: Hardware, Software, Regulations, Standardization and Organization. Our focus in this dissertation is on software issues. We propose new methods to generate a reliable context in a pervasive information system that has high rates of new measurements over time using data aggregation and data fusion. Different aggregation and fusion techniques can be applied depending on the types of sensed data and autonomous processing within the fusion step.The goal of this research is to produce a high confidence level in the generated context for remote monitoring of patients. Reliable contextual information of remotely monitored patients can prevent hazardous situations by recognizing emergency situations in home-based care. However, it is difficult to achieve a high confidence level of contextual information for several reasons. First, the pieces of information obtained from multi-sensors have different degrees of uncertainty. Second, generated contexts can be conflicting even though they are acquired by simultaneous operations. And last, context reasoning over time is difficult because of unpredictable temporal changes in sensory information. In particular, some types of contextual information are more important than others in home-based care. The weight of this information may change, due to the aggregation of the various sensors (evidence) and the variation of the values of the various sensors (evidence) over time. This causes difficulty in defining the absolute weight of the evidence in order to obtain the correct decision making.In this dissertation, we propose an evidential fusion process as a context reasoning method based on the defined context classification and state-space based context modeling. First, the context reasoning method processes sensed data with an evidential form based on Dezert-Smarandache Theory (DSmT). The DSmT approach reduces ambiguous or conflicting contextual information in multi-sensor networks. Second, we deal with dynamic metrics such as preference, temporal consistency, and relation-dependency of the context using Autonomous Learning Process (ALP) and Temporal Belief Filtering (TBF) in order to improve the confidence level of contextual information that makes a correct decision about the situation of the patient. And last, we deal with both relative and individual importance of the evidence to obtain an optimal weight of the evidence. We then apply dynamic weights of the evidence into Dynamic Evidential Network (DEN) in order to improve the confidence level of the context and to understand the emergency progress of the patient in home-based care.Finally, we compare the Evidential Fusion Process on DSmT with traditional fusion processes such as Bayesian Networks (BNs), Dempster-Shafer Theory (DST), and Dynamic Bayesian Networks (DBNs). This comparison makes us understand the uncertainty analysis in decision-making by distinguishing sensor reading errors (i.e., false alarm) from new sensor activations or deactivations, and shows the improvement of our proposed method compared to the others.The main contributions of the proposed context reasoning method under uncertainty based on evidential fusion networks are: 1) Reducing the conflicting mass in uncertainty level and improving the confidence level by adapting the DSmT, 2) Distinguishing the sensor reading error from new sensor activations or deactivations by considering the ALP and the TBF algorithm, and 3) Representing optimal weights of the evidence by applying the normalized weighting technique into related context attributes. These advantages help to make correct decisions about the situation of the patient in home-based care.