A Wearable Data Collection Platform with Smart Annotation Capabilities
Abstract
Remotely tracking users’ activity and physiology can help on disease treatment and health monitoring. For example, in nutrition management, tracking food intake helps on weight control. However, to train tracking algorithms, annotated data is needed which is typically obtained manually. Users’ manual annotation is challenging as it’s affected by factors such as recall bias and may become a burden, causing them to stop annotating. Automatic approaches exist, but they may not personalize to individual users, resulting in inaccurate annotations. Therefore, personal pattern identification and adaptation are needed to achieve a satisfactory annotation process. We present a system capable of personal patterns’ identification and intelligent data annotation for accurate personal monitoring without burdening the user.
Subject
Data collectiondata annotation
machine learning
deep learning
neural networks
diet monitoring
Citation
Solis Castilla, Roger Fernando (2019). A Wearable Data Collection Platform with Smart Annotation Capabilities. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /186320.