projects

Student Projects

  • 2016 SH project on “SensorCube: An end-to-end framework for conducting research via mobile sensing”
  • 2015 SH project on “Capturing social cues in face-to-face conversation for the visually-impaired”
    • Student: Lauren Murray
    • This work will be presented at the international workshop of BodySenseUX@UbiComp 2016, Germany, in September 2016.
    • Paper is available at: pdf
  • 2015 SH Project on “Pedestrian fear detection using gait analysis on smartphones”
  • 2014 SH project on “Behaviour change intervention for problematic mobile phone use in social situations”
    • Student: Thomas Eddie
    • The work has been presented at the International Workshop on Intelligent Attention Management on Mobile Devices, cohosted with MobileHCI 2015. Copenhagen, Denmark, 24th August 2015.
    • Paper is available at: pdf
  • 2014 SH project on “Development of a platform for high throughput genomic analysis”, collaboration with Dr. Silvia Parachini in the school of Medicine at the University of St Andrews
    • Student: Valentin Tunev
    • The software is accessible at github

Potential PhD Projects

Pervasive systems must offer an open, extensible, and evolving portfolio of services which integrate sensor data from a diverse range of sources. The core challenge is to provide appropriate and consistent adaptive behaviours for these services in the face of huge volumes of sensor data exhibiting varying degrees of precision, accuracy and dynamism. My research interest centres around how to understand and program such sensor data and make them useful for high-level applications. More particularly, I am interested in the following topics:

Activity recognition

Sensor data comprehension and interpretation relies on the ability to infer human activities that give rise to a particular data stream. Activity recognition is a challenging research topic, far beyond traditional classification problems, which is mainly due to both the variety, heterogeneous, and noise of sensor data and the complexity and ambiguity of human behaviours. Machine learning and data mining techniques have been attempted to solve small-scale activity recognition problems in the past few years, but there are still lots of open areas that have not been (at least sufficiently) explored.
Most machine learning and data mining techniques rely on training data; however, training data in pervasive environments are often difficult to collect or are poorly annotated. The main research question is how a new activity recognition technique can transfer the knowledge learned on a limited number of users to a much larger pool of users and adapt the learning in a long term. How can the technique leverage the use of open source knowledge like Wikipedia or WordNet, open social media data like twitter or Four-square, or more generally crowd-sourcing?

Uncertainty management and programming

Uncertainty is a pervasive factor in many sensor-enabled infrastructures, compromising the integrity of the resultant dataset. Incorrect calibration, deteriorating performance over time, or the omnipresent noise within the physical environment, may all reduce the quality and provenance of the data. Imperfect sensor data may result in incorrect inference of real-world phenomena, including human activity recognition for example, thereby reducing the perceived quality of service and experience. Thus a model for handling uncertainty is essential for robust and human-centric ambient intelligence services. The research question is how to address this critical issue of uncertainty in ambient intelligence, with an emphasis on strategies for detecting, resolving, and programming against uncertainty.

Behaviour-aware computing

Behaviour-aware computing is a relative new research topic, immediately following context- and activity-aware computing. It is not only about inferring or predicting human behaviours, but more about adapting to user behaviours and changing their behaviours. This is an interdisciplinary research, often involving how to apply theories and models in psychology to designing a feedback system. Earlier attempts reside in shaping pro-health and -environmental behaviours including diet, fitness, green transportation, energy consumption etc. The main research challenge is sustainability – how to sustain user’s “good” behaviour in a long-term period.

Past MSc Projects

  • Uncovering biodegradation impacts in contaminated soil using statistical data analysis
    The goal of this project is to investigate new approaches for intelligent analysis of the data with particular focus on two aspects:
    1. The difference in reaction rates of positional isomers will be studied to establish the importance of molecular shape and electron distribution in the biodegradation mechanisms. Using a particular family of positional isomers (that of dimethyl naphthalene), the conditional dependency and mutual information relationship between the studied parameters will be examined using Shannon?s information theory. Correlation rules between molecular structure and reaction rates will be generated by higher-level techniques like decision trees or genetic algorithms.
    2. The relationship between the logical order of elution of compounds in a GCxGC and the reactivity to biodegradation will be investigated. Elution of compounds in GCxGC set-up was shown to be predictable based on a series of parameters, the linear free energy descriptors, that relate to the size and the polarity of the compounds. These descriptors have been used to help predict the evaporation and water washing rates of hydrocarbons using GCxGC chromatograms but never been related to biological reactivity. We will experiment both linear and nonlinear (i.e. artificial neural network and support vector machine) techniques to investigate the relationships between these descriptors and the biological reactivity so as to find out whether they can be used to describe biological models and/or whether they need extended to other molecule specific parameters such as the ones identified in the isomers study.
  • development of computer algorithms for visual-to-haptic image translation
    The goal of this project is the development of computer algorithms for visual-to-haptic image translation. While previous devices have used haptic feedback for communicating form, they rely on existing haptic maps to communicate simple structures. Conversely, research into the perception of haptic form has used physical haptic surfaces. Understanding how artificial vibration feedback affects the haptic perception of form is essential. The algorithms to be developed will be tailored to allow for the meaningful haptic representation of image information, using image processing strategies to explicitly encode important image features. We shall actively engage with the public, especially those with visual impairments, in order to refine our algorithms.

Funded Projects

  • (Co-I) Describing multiple contaminants biodegradation in soil using comprehensive two-dimensional gas chromatography coupled with intelligent data analysis. Scottish Crucible Award, 2013
    • Start from 1st Nov 2013, 15 months
    • Collaborate with University of Glasgow, Heriot Watt University
  • (Co-I) Perceiving Pictures Through Touch: A Haptic Interface for Communicating Form. Scottish Crucible Award, 2013
    • Start from 4th Nov 2013, 14 months
    • Collaborate with University of Stirling, Edinburgh University, University of Glasgow

Participated EU Projects

Past Research Projects

  • 2009-present “Autonomic Sensor Communities” in CLARITY Project
  • 2008-2009 “Situation Lattices” to systematically study the semantics of information flow in pervasive computing systems.
  • 2007 “Location in Pervasive Computing”, a unified semantic spatial model to represent all types of location information and their spatial relationships.
  • 2006 “Ontology Models in Pervasive Computing”, an ontology-based context model for pervasive computing systems.