Multimedia Data Engineering Laboratory

Japanese
Affective Engineering for Multimedia Data

Retrieval, presentation, and analysis of multimedia data are studied based on their impressions.

  1. Cross-media data retrieval based on impression
  2. Retrieval of multimedia data including pictures, videos, and sound and music clips is studied. Especially, different types of media data are focused on.

    We use the impression spaces, each of which is for a media type. Positions in the space represnt impressions of media data. As impression spaces are easily associated with each other, different media data are mutually retrieved. For example, music clips suitable for a picture are obtained. It is also possible to obtain pictures suitable for a music clips. Any multimedi data can be the targets of retrieval by estimating the positions of multimedia data in an impression space from their feature values.

  3. Synchronization of music and video clips for invoking deep impression
  4. Different types of media data, e.g., video and music clips, are tried to be presented with synchronization. This improves the structural harmony of different types of media data. This may result in invoking deep impression.

  5. Affective multimedia data retrieval considering feeling and personality
  6. Behavior of the affective retrieval is changed according to feeling and personality. That is, the retrieval results varies according to feeling happy or sad, and/or according to a person.

Copyright © 2008-2015 Multimedia Data Engineering Lab. All Rights Reserved.