Ms Anastassia Angelopoulou, UoA 11, ECS
Dr Anastassia Angelopoulou received her BSc in Graphic Arts Technology (1997) from the Technological Educational Institute of Athens, her BSc in Mathematics (2009) from the Open University and both her MSc in Multimedia (1999) and PhD in Computer Science (2012) from the University of Westminster. For the last fifteen years she has been leading research and industrial projects related to manmachine interfaces and pervasive gaming. She is currently a Senior Lecturer at the Faculty of Science and Technology and leads the Computer Games Development BSc Honours course. She has published more than 40 peer reviewed articles in journals and conferences including Neural Networks, IEICE Transactions, ECCV, IJCNN, ICCV, HCI, WCCI and ACM. She has also served as a programme committee member and reviewer in IET and Elsevier Journals, and IEEE and ACM conferences.
Current projects include:
- multi-platform augmented video gaming in theatres (Southwark Playhouse)
- mobile game apps in museums (Sutton Hoo)
- AR-interfaces (Victoria & Albert museum)
- online games (City of Westminster Archives, Thames Water)
This research is motivated by a desire to provide a self-organising framework for the unsupervised model acquisition and tracking of non-rigid objects. In any learning framework, the initialisation of the shapes is very crucial. This is a difficult task with many approaches using either hand-labelling training samples or pre-trained classifiers, both of which are instances of supervised learning. In this research, we develop an automatic, complete, unsupervised 2D hand-gesture model acquisition system using topological structures with no fixed dimensionality, and descriptors with no a priori knowledge of the network size. We first develop a framework for building non-rigid shapes using the growth mechanism of the self-organising maps. Then we define an optimal number of nodes without overfitting or underfitting the network, based on the knowledge obtained from information theoretic considerations. The modified network consists of descriptors obtained from a spatial transformation of the network, an automatic criterion for maximum node growth and local features for object tracking. This model is used to represent motion in image sequences by initialising a suitable segmentation. The work has applications in the online tracking and segmentation of previously unmodelled objects and their incorporation in other environments such as augmented reality.