Content-based Segmentation for Audio-Visual Information Retrieval

Speaker:
Petra Galuščáková
Abstract:
In this talk we deal with Information Retrieval from audio-visual recordings. Such recordings are often long and a user may want to know the exact relevant passage of the recording. Therefore, the recordings are automatically divided into smaller parts, on which we apply standard retrieval techniques. We  experiment with various methods for segmentation of the recordings into shorter segments which are then used in a standard retrieval setup to search for relevant passages. The main focus will be on machine-learning based approaches utilizing the content of the recordings. We will describe experiments performed in two shared tasks of MediaEval Benchmark: Similar Segments in Social Speech and Search and Hyperlinking.
Length:
01:03:35
Date:
03/03/2014
views: 804

Images:
Preview of img-004.jpg
Image img-004.jpg
Preview of img-026.jpg
Image img-026.jpg
Preview of img-037.jpg
Image img-037.jpg
Attachments: (video, slides, etc.)
60MB
226 downloads
260MB
227 downloads
60MB
215 downloads