Loading...
Please wait, while we are loading the content...
Similar Documents
SceneCabinet / Live ! : Realtime Generation of Semantic Metadata Combining Media Analysis and Speech Interface Technologies
| Content Provider | Semantic Scholar |
|---|---|
| Author | Kuwano, Hidetaka Kon'ya, Yuko Yamada, Tomokazu Kawazoe, Katsuhiko |
| Copyright Year | 2004 |
| Abstract | Reducing the cost of generating metadata would allow more broadcast contents to be transmitted with advanced TV viewing services available through the use of metadata. In this article, we describe SceneCabinet/Live!, a system that generates scene-based semantic metadata for video content in real time by combining media analysis and user interface technologies. The system provides an intuitive operation interface that lets the operator easily generate metadata by only speaking about the content while watching the replayed video. No keyboard input is required for this operation. Scene title, synopsis, and keywords can be obtained using natural language processing based on speech recognition results. Our speech recognition method obtains almost errorless results because it uses specific grammatical rules matched to the genre of video content. In an experiment, for a live baseball program broadcast, we confirmed that semantic metadata concerning scenes of home runs and skilful play could be generated in real time without any delay. SceneCabinet/Live!: Realtime Generation of Semantic Metadata Combining Media Analysis and Speech Interface Technologies Special Feature † NTT Cyber Solutions Laboratories Yokosuka-shi, 239-0947 Japan E-mail: kuwano.hidetaka@lab.ntt.co.jp |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://ntt-review.jp/archive/ntttechnical.php?contents=ntr200508040.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |