Research Overview

Digital Cinema and Restoration

In the Digital Cinema stream our work in Digital Motion Picture Restoration is well known and perhaps the earliest in the area, we have worked on many EU Projects involving restoration (AURORA, BRAVA, PRESTOSPACE) and have associations with archives around Europe. We are also involved in the creation of new cinema post-production tools and our motion estimation technology (in collaboration with The Foundry) is used around the world for film special effects in movies like Matrix, Lord of the Rings, X-Men, Spider Man, Harry Potter and so on.

We have just finished working on the EU project i3dpost, a collaboration between TCD, uni Surrey, Uni Thessonaliki and Buf (a post production house in France), QuanticDream (Paris based games developer). The idea is to exploit multiview film recording for new post-production techniques and more robust editing tools.

Information Retrieval

We have been involved with projects in Information Retrieval since 1999. We coordinated MOUMIR and worked on Enterprise Ireland projects involving the automated creation of summaries for sports video: snooker, tennis, cricket. We have also established a unique connection with the Dept. of Psychology regarding the diagnosis of Dyslexia using new video retrieval and motion analysis technology ]. Our ethos in the development of retrieval technology has been that "retrieval is not a science until the user context is properly acknowledged". To this end we are involved with an ongoing collaboration with the University of the West Indies, Trinidad and Tobago, to exploit retrieval technology in reducing the infrastructure required for Distance Learning. Projects involving retrieval for parsing of video from sports coaching sessions are also underway in this collaboration. Recently we have started work with the Royal College of Surgeons in Ireland to assist with automated editing and enhancement of video from anatomy practicals.

Human Speech Communication

Wherever humans communicate using the medium of speech, whether amongst themselves or to computers, we use engineering expertise to manipulate, enhance, or fix the speech. A recurring theme in this research is that speech is bimodal in nature. By exploiting both audio and visual cues, we can improve machine understanding of speech. We have ongoing projects in:

  • Enhancing speech intelligibility for digital heading aids and hearing-impaired listeners
  • Assessing speech quality in internet-based communication
  • Speaker verification for biometrics, and the impact of vocal ageing on biometric systems
  • Audio-visual speech recognition
  • Recognising emotion from speech
Page last modified on May 16, 2012