Weakly Supervised Deep Learning for Human Activity Recognition

Thumbnail Image
An, Sungtae
Inan, Omer T.
Rehg, James M.
Associated Organization(s)
Organizational Unit
Organizational Unit
Supplementary to
Human activity recognition (HAR) using wearable sensors and machine learning algorithms is an emerging capability in domains including but not limited to healthcare and ergometric analysis of populations by providing context to physiological measures from wearable sensors during natural daily living activities. Despite the success of deep supervised models in recent years, obtaining a fully labeled HAR dataset is often challenging due to the high cost and workforce associated with labeling. In this dissertation, I address the challenge with the following contributions. First, I present the bilateral domain adaptation problem in HAR for the first time and propose AdaptNet, a semi-supervised deep translation network. AdaptNet enables information fusion of two different data domains using both unlabeled and labeled data. Next, I present a novel framework, DynaLAP, a semi-supervised variational recurrent neural network with a dynamic prior distribution, to perform activity recognition in fixed routes and protocols. DynaLAP implicitly exploits the information about the environment to enhance HAR in fixed protocols such as military and athletic training with few labeled subject data. Then, I present DualCPC, self-supervised pre-training with the contrastive predictive coding framework using a tri-axial accelerometer signal and corresponding physiological variable measurements such as instantaneous oxygen uptake (VO2) during activities performed. Utilizing an auxiliary physiological variable in training time only, the DualCPC pre-trained model outperformed the baseline models across different numbers of labeled training data available. Finally, I conclude the dissertation by discussing the potential future work and extensions.
Date Issued
Resource Type
Resource Subtype
Rights Statement
Rights URI