Recognizing Human Activities from Physical and Virtual Worlds

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "Recognizing Human Activities from Physical and Virtual Worlds"

By

Mr. Hao HU


Abstract

Recognizing the activities underlying human actions has been an extensive 
research topic since early 1980s, where researchers usually focus on 
understanding the activities of human in the physical world. Recently, with the 
advent of OSNs (online social networks), more and more online social activities 
also start to emerge. In this proposal, we aim to provide solutions for 
recognizing human activities both in the physical world and in the virtual 
world. We start by surveying related works in these two areas and then study 
some specific challenges which are important to deploy these activity 
recognition systems in the real world.

In this thesis, we investigate a number of specific problems we need to tackle 
when deploying activity recognition techniques in the real-world. We first 
analyze how to recognize multiple activities in the physical world environment, 
especially when such activities have concurrent and interleaving relationships. 
Next, we extend such a framework into the virtual world, by exploiting the 
relatedness of search queries to activities with interleaving relationships and 
propose a context-aware query classification algorithm. Secondly, we study the 
problem of recognizing abnormal activities. These abnormal activities are rare 
to happen and it is difficult to collect enough training data about them. We 
develop an algorithm based on the Hierarchical Dirichlet Process and the 
one-class Support Vector Machine to recognize abnormal activities when the 
training data is scarce. Finally, when we need to deploy the activity 
recognition systems in the real-world, it is impractical for us to collect 
enough training data for different activity recognition scenarios, especially 
when we need to collect training data for different persons and even for 
different actions. To solve this problem, we've developed a transfer 
learning-based activity recognition framework which borrows useful information 
from previously collected and learned activity recognition domains and then 
re-use such information into the new target activity recognition domain.

Furthermore, we've conducted extensive experiments to demonstrate the 
effectiveness of our proposed approaches on real-world datasets collected from 
smart homes or sensor environments. We've also shown that our context-aware 
query classification algorithm could outperform state-of-the-art query 
classification approaches on real-world query engine search logs. At the end of 
this thesis, we discuss some possible directions and problems for future work 
and extensions.


Date:			Monday, 26 November 2012

Time:			10:00am - 12:00noon

Venue:			Room 3402
 			Lifts 17/18

Chairman:		Prof. Yu-Hsing Wang (CIVL)

Committee Members:	Prof. Qiang Yang (Supervisor)
 			Prof. Raymond Wong
 			Prof. Dit-Yan Yeung
 			Prof. Ke Yi
 			Prof. Rong Zheng (ISOM)
                        Prof. Xing Xie (Microsoft Research Asia, China)


**** ALL are Welcome ****