Lyndon Kennedy
Dept. of Electrical Engineering
Columbia University
500 W. 120th Street, 13th Floor
New York, NY 10027
Office: 716 CEPSR
Lab: 7LE3 CEPSR
Phone: 212 854-0513
Fax: 212 932-9421
Email: lyndon@ee.columbia.edu
Ph.D. Candidate, Digital Video and Multimedia Lab
Department of Electrical Engineering, Columbia University
I am a Ph.D. candidate and graduate research assistant in the the Department of Electrical Engineering at Columbia University. I hold B.S. (2003), M.S. (2005), and M.Phil. (2008) degrees, all from Columbia University in Electrical Engineering. I conduct research in the Digital Video and Multimedia Lab under the direction of Prof. Shih-Fu Chang. I am currently investigating techniques for automatically indexing and searching digital image and video collections.
My research has received several awards and honors. I was co-author of a paper awarded the IBM Student Paper Award at ICIP in 2004. In 2006, I was co-author of a paper nominated for the Best Paper Award at ACM Multimedia. Most recently, in October 2007, I was invited as part of a group of 8 top senior Ph.D. students in the multimedia field to participate in the IBM Watson Emerging Leaders in Multimedia Workshop.
I spent the summer of 2007 working as an intern at Yahoo! Research Berkeley with Mor Naaman. Before beginning my work in the DVMM Lab (in 2003 and 2004), I worked on the Mapping Meetings project under Prof. Dan Ellis and on a Robotic In-Vivo Imaging Device under Prof. Peter Allen.
I am interested in methods for enabling semantic search over multimedia collections. Many current solutions to this problem (particularly, web image search engines) rely primarly on textual metadata. The approaches that I am pursuing aim to improve upon these techniques by also leveraging the audio/visual content of the documents. Recently, I have been exploring the ways in which multimedia documents are re-used by many authors and how we can leverage the resulting emergent internal structures of multimedia collections to improve search, summarization, and exploration. In particular, content-based approaches, which can sometimes seem inadequate on their own, can be revisited and re-imagined in these new multiple-author environments (such as on the web or in social media sharing sites) and yield novel and robust applications.
In the past few years, I have been a frequent participant in the NIST TRECVID video retrieval benchmark and have designed a few high-performing video retrieval systems. I also had an active role in the development of the LSCOM Lexicon and Annotations, a first-of-its-kind collection of hundreds of visual concepts annotated over a large digital video collection, which has been downloaded by hundreds of researchers worldwide.