A novel, interactive Android app has been developed that monitors the health of type 2 diabetic patients in real-time, providing patients and their physicians with real-time feedback on all relevant parameters of diabetes. The app includes modules for recording carbohydrate intake and blood glucose; for reminding patients about the need to take medications on schedule; and for tracking physical activity, using movement data via Bluetooth from a pair of wearable insole devices. Two machine learning models were developed to detect seven physical activities: sitting, standing, walking, running, stair ascent, stair descent and use of elliptical trainers. The SVM and decision tree models produced an average accuracy of 85% for these seven activities. The decision tree model is implemented in an app that classifies human activity in real-time.
| Identifer | oai:union.ndltd.org:ETSU/oai:dc.etsu.edu:etd-5113 |
| Date | 01 December 2019 |
| Creators | Chowdhury, Nusrat |
| Publisher | Digital Commons @ East Tennessee State University |
| Source Sets | East Tennessee State University |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Electronic Theses and Dissertations |
| Rights | Copyright by the authors. |
Page generated in 0.0018 seconds