Identifying The Usage Anomalies For ECG-Based Healthcare Body Sensor Networks

This thesis is looking into the dependability of a Electrocardiogram(ECG) based Healthcare Body Sensor Network system (HC-BSNs). For these type of devices, the dependability is not only depending on the devices themselves, but also heavily depending on how the devices are used. Existing literature has identified that there are around 4% of usage issues when existing ECG devices are used by professionals. The rate of usage issue will not be better for the ECG-Based HC-BSNs as these devices are more likely to be used by untrained people. Subsequently, it is with paramount importance to address the usage issues so that the overall dependability of the ECG-Based HC-BSNs can be assured. Our approach to address the usage issue is to detect the usage-related anomaly, which is contained in the captured signal when erroneous usage is made, and identify the cause to the usage-related anomaly automatically and without human intervention. By doing this, the user can be prompted with clearer and accurate correction instruction. Subsequently, the usage issues can be well corrected by the user. Based on the above concept, in this thesis, we have studied the anomalous signals which can be caused by the usage issues. Two methodologies, names as AID and FFNAID, have been proposed and evaluated to detect the usage-related anomalies. We have also studied how each usage issue can affect the signals on a mote, and we use the knowledge learnt from the study to propose a methodology, named as ACLP, to identify the root cause to the usage-related anomaly. All these methodologies are fully automated and does not require any human intervention once they are deployed. The evaluations have also shown the effectiveness of these methodologies.