Digital Library

Abstract: In this talk, I give an overview of recent and ongoing work at ORL in the area of multimedia information retrieval (MMIR). Our lab has a strong background in high-bandwidth multimedia systems, and it was in fact the real problem of dealing with our increasingly large archives of video mail which motivated our first MMIR project, Video Mail Retrieval (VMR), we conducted jointly with the speech recognition group of the Cambridge University Engineering Department. The first version of VMR enabled the retrieval of video documents based on scanning the audio track for keywords, and the final version used pre-computed phoneme lattices for open-vocabulary and speaker-independent retrieval. The system was also applied to the retrieval of items of interest from an archive of several months worth of television broadcast news. The VMR project finished this year, but we are now ramping up a follow-on MMIR project which will consider indexing on image content (using both still-image and, in the case of video, inter-frame analyses) as well as further improvements to audio indexing (using, for example, audio classification techniques in an initial phase). The goal is a system which supports multi-modal queries on large heterogenous multimedia collections, including those drawn dynamically from the web.

Speaker's Biography: Kenneth R. Wood is a research scientist at Oracle Olivetti Research Laboratory, the jointly funded research laboratory of the Oracle and Olivetti corporations. His interests include multimedia information retrieval, mobile and ubiquitous computing, concurrency theory, and applied formal methods. Before joining ORL, he worked as a member of scientific staff at Bell-Northern Research (now Nortel Technology) and taught for several years at Oxford University. He holds the AB degree in Applied Mathematics from Harvard University and the MSc and DPhil degrees in Computation from Oxford University.

Presentation: This presentation is in the form of an audio file and a both HTML and PDF files containing the author's slides. You can experience this presentation as follows:

• Click on this Start First Audio link initially, then Start Second Audio for the remainder.
• Click on this Start HTML link (while the first audio file is loading) OR
• Click on this Start PDF link (while the audio file is loading).
• Use the "Next Page" button to advance the slides as the talk progresses.