PhD research

I am part-time PhD student, under Prof. Claude Sammut (CSE) and Dr. Fang Chen (NICTA).

Head interaction user experiment

Multimodal user interfaces (MMUI) allow humans to interact with machines using a variety of communication channels such as speech or gesture, and conversely machines to convey information to humans using modalities such as visual graphics and text, or synthesised speech. There is a general agreement on the benefits of multimodal user interaction, i.e. it provides more natural, intuitive and cognitively efficient interaction, but current research falls short of methods to design and evaluate related systems. The main objective of this research is to advance the understanding of how do humans interact multimodally, and to provide a methodology for the creation of MMUI, taking into account the individual variability across users, and to automate some of the processes in order to allow industrial deployment.

Adaptive multimodal user interaction, my topic, refers to adapting a human computer interface's input and output processes to a particular person's productions using several modalities. Detecting specific multimodal interaction patterns (MIP) utilised by that person can improve the automatic recognition of the user's input, as well as help convey information back to that user in a more appropriate way. For example, if the person tends to point at objects before uttering the command he or she wants to perform on these targets, the system may exhibit a similar sequential behaviour. But such mimicking may not be desirable all the time, if at all. This research aims to address fundamental issues such as the correlation between semantic content, temporal arrangements and spatial characteristics of MIP. Based on such findings, suitable strategies can be predicted for input fusion and output generation processes. A strong commitment to industry-deployable results underpins this research; hence it will explore the limitations of traditional user-centred design (UCD) methods for MMUI design, especially the lack of evaluation metrics for such systems. It will propose enhancements leading to the elicitation, capture, analysis and exploitation of multimodal interaction patterns from human subjects. Three major research questions will be addressed by this research:

  1. Determination of relevant multimodal interaction patterns, in terms of semantic content, types of modalities and temporal relationships;
  2. Prediction of optimal input fusion and output generation schemas according to MIP;
  3. Definition of evaluation metrics for MMUI systems, in a UCD context.

Professional research

I am also a senior research engineer within National ICT Australia (NICTA), located at the Australian technology park, Sydney. I am working within the Multimodal User Interaction team, on the Smart Roads and Tranport, Multimodal User Interfaces (STaR-UI) project.

Publications

contact me