Abstract

Abstract: Driven by the persuasiveness of sensing technologies in our everyday devices, activity recognition has been widely adopted in many applications, from smart home, personal healthcare, human-computer interaction, to name a few examples. The key enabler to these application is the ability to accurately recognise people’s activities. Activity recognition techniques, especially those based on deep neural networks, have made significant progress on learning complex correlations between sensor data and activity classes. However, they often rely on a large amount of well-annotated training data, typically time- and effort-expensive, if not feasible at all, to acquire. To tackle this challenge, we have designed various unsupervised domain adaptation techniques to enable transferring the activity knowledge from one environment to many other environments. In this talk, we will introduce 3 of our recent developed techniques: (1) knowledge-driven ensemble learning, (2) variational Autoencoder-based feature space alignment, and (3) bi-directional generative adversarial networks-based feature space transfer.

Bio: Dr. Juan Ye is a Senior Lecturer in the School of Computer Science at the University of St Andrews. She received a Bachelor’s and Master’s degree from Wuhan University, China, in 2002 and 2005 respectively, and a PhD in Computer Science from University College Dublin, Ireland, in 2009. Her research interest centers on human activity recognition, with a particular focus on domain adaptation, continual learning, and unsupervised learning.