Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 55628 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Jahed, Mehran
- Abstract:
- Wireless area body networks (WBANs) are a set of sensors designed to monitor an individual’s health remotely. These networks are required to transfer the data properly to the healthcare professional. Since there is no guarantee that no faults might occur, these networks should have fault tolerance capability such that they can maintain their main function properly. In this thesis, a method for fault tolerance in WBANs is proposed.In this research, to test the proposed method, a case study on sleep apnea disorder is selected for monitoring patients who experience sleep apnea. To this end, the Apnea-ECG dataset from the PhysioNet repository was considered which contains five different types of sensors that are placed on the subjects. These sensors are utilized to estimate the performance of comprehensive respiratory behavior and sleep apnea disease, namely single-lead ECG signals, chest and abdominal respiratory effort through inductive plethysmography, nasal airflow through nasal thermistor, and blood oxygen saturation through pulse oximetry sensor.In the proposed method for fault tolerance, the cooperation of sensors and intelligent inference is used so that the sleep apnea monitoring system may continue to work correctly in the presence of a fault. The cooperation of sensors in this study means that despite the fault in the data of one of the sensors, by referring to the correlation between the data of the sensors from the physiological point of view, the presence of the fault is verified and the data of remaning sensors become the basis for estimating the behavior of the respiratory system.In this study, the motion artifact of the respiratory effort sensor is selected as a candidate fault as its behavior cannot be easily distinguished from respiratory signals. In this regard, the performance of the proposed method is carried out in two phases, namely an overall evaluation or the first phase, and an evaluation based on the data of an individual or the second phase. Furthermore, each phase includes three scenarios where the method of injecting the fault would differ. In the first scenario, after separating the training and test data, the fault is injected into the test data. In the second scenario, first, the fault is injected into a percentage of the data and then the training and test data are separated. In the third scenario, similar to the first scenario, the training and test data are separated. However, in the third scenario unlike the first one, fault injection is applied to the training and test data separately. Different criteria were used in the evaluation process, inclusive of accuracy, precision, recall, F1 score, specificity, receiving characteristic operating curve, and the area under the curve. Based on this analysis, the third scenario was chosen as the most appropriate, as in this scenario, the percentage of independent fault injection for training and testing data was considered comprehensively. Promising results were obtained in the presence of a fault, where the overall evaluation accuracy, namely first phase of the third scenario of the fault tolerant system was 96.24% and the average individual evaluation, namely the second phase of the third scenario accuracy of the fault tolerant system was 96.02%.
- Keywords:
- Fault Tolerance ; Wireless Body Area Network ; Apnea ; Machine Learning ; Apnea Detection ; Electromyogram Signal ; Sleep Disorder Breathing (SDB) ; Cooperation System
Digital Object List
-
محتواي کتاب
- view
Bookmark