2004 PhD Scholars

   

Alban RrustemiAlban Rrustemi
University of Cambridge, United Kingdom

Supervisor: Dr. Simon Moore
Microsoft Research supervisor: Dr. Ken Wood

Research title: Dense wired sensor networks

Research summary: Computing fabrics could be constructed from small elements (for example, 1.5mm x 1.5mm chips) woven into clothing or embedded into the structure of some device. Communication amongst these elements and other components would allow useful systems to be constructed. Reconfiguration approaches will address the void between conventional architectures of microcontrollers + software at one extreme to field programmable gate microcontrollers + arrays (FPGAs) at the other. To enable prototypes to be made, Alban will draw on existing research and industrial expertise in the areas of sensors, chip packaging and power sources.

Anna RitchieAnna Ritchie
University of Cambridge, United Kingdom

Supervisor: Dr. Simone Teufel
Microsoft Research supervisor: Prof. Stephen Robertson

Research title: Combining term-based and citation-based methods for enhanced information retrieval

Research summary: The aim of the project is to combine citation information with traditional term based IR information sources for more sophisticated search. In which form citation information is to be included is itself an object of research: one could use the citations themselves, the citation anchor text as context, and possibly discourse structure information of the segment where the citation occurs.

Dana N. XuDana N. Xu
University of Cambridge, United Kingdom

Supervisor: Prof. Alan Mycroft
Microsoft Research supervisor: Dr. Simon Peyton-Jones

Research title: Software tools for secure component programming

Research summary: With our ever-growing reliance on software systems – occasionally for life-critical situations, it is no longer safe to rely solely on the informal assurances of software testing. Dana proposes a new framework for software systems to be safely (and securely) built from software components. Several major verification techniques have been advocated over the last two decades, including model-checking, static analysis and advanced type theory. However, little attention has been paid to applying these techniques to component-based programming. In this project, Dana plans to explore both the foundations for secure component integration and practical tools to support its development.

Greg HaleGreg Hale
University of York, United Kingdom

Supervisor: Prof. Andrew Monk
Microsoft Research supervisor: Dr. Ken Wood

Research title: Qualitative and quantitative studies of ‘fun’ with interactive feature film based entertainment via mobile phones and Web sites

Research summary: This research project is an investigation of entertainment content experiences, focused particularly on movies and narrative based mobile games/ interactive stories. The objective is to create an integrated psychological framework of these content experiences, which in turn can inform content design. Work completed. The first empirical study involved a systematic qualitative investigation of people’s responses to a short film, using interview data. This work is being integrated into the relevant literature from psychology, human-computer interaction and cognitive approaches to film, around a specific focus of schemas. The work has resulted in two conference papers, presented at international conferences. Work remaining. The integrated framework will be used to analyse three successful movies from the same genre, a ‘systems’ investigation. The flow of schematically structured experience events in the films will be logged and mapped onto the psychological framework. A short film will be then created using the framework and tested on viewers, with interview data being gathered. Finally, the framework will be tested for robustness in a different context by analysing either an existing mobile game or interactive story.

Julia LasserreJulia Lasserre
University of Cambridge, United Kingdom

Supervisor: Dr. Roberto Cipola
Microsoft Research supervisor: Prof. Christopher Bishop

Research title: Bayesian object recognition using weakly labelled data

Research summary: This PhD will focus on the problem of object recognition using weakly labelled data, it will build on recent developments in probabilistic modelling and Bayesian inference. A typical task, for example, will be to learn about and recognise say cars given only a pile of images containing cars and a pile not containing cars. At some level this is very feasible, but a general solution will be very challenging to find. A central point of visual perception is the classical problem of invariant object recognition: different appearances of an object can be seen as equivalent, but with changes in position, illumination, distortions, or partial occlusion by other objects. It can then take account of prior knowledge to do with geometrical transformations of objects, but can also to some extent learn invariant features from the training data. One approach which may prove very helpful involves capturing multiple images of the same object from different orientations and viewpoints. Prior knowledge of the image generation process including projective geometry can be used to exploit known geometrical transformations and invariants.

Mathieu VerbaereMathieu Verbaere
University of Oxford, United Kingdom

Supervisor: Prof. Oege de Moor

Research title: An extensible toolkit for refactoring

Research summary: To enable developers to author their own refactoring transformations. Existing tools offer only a fixed menu of refactorings. Furthermore, even simple refactorings like ‘extract method’ are often incorrectly implemented, because the implementation does only syntactic and no semantic analysis. Mathieu aims to construct an extensible framework for easy and correct implementation of refactoring transformations at several levels: a set of static analyses that frequently occur in refactoring; an API for using these analyses, and applying the corresponding transformations to the source (retaining layout and comments).

Maurice FallonMaurice Fallon
University of Cambridge, United Kingdom

Supervisor: Dr. Simon Godsill
Microsoft Research supervisor: Prof. Andrew Blake

Research title: Multi-channel audio source localisation and tracking

Research summary: In this project Maurice will study signal processing methods for enhancing audio signals obtained from microphone arrays. The methods will aim initially to improve on the current state of the art in echo cancellation, then develop into the areas of source separation and localisation. The approach will be based on time-frequency models of the audio signals and the inclusion of Bayesian prior information about coefficients across time and frequency in order to aid the enhancement process. A selection of methodologies will be pursued, including variational methods, particle filters and fast approximations to these.