Introduction - Multimedia Database Retrieval: Technology and Applications - page 14

allows the description of the uniqueness of gestures for recognition, in terms of
trajectories on the map. The hidden Markov model is then utilized for trajectory
analysis. Within the system, the dance sequence of the student can be segmented
online and cross-referenced against a library of gestural components performed by
the teacher. This facilitates the assessment of the student dance, as well as provides
visual feedback for effective training.

1.4 Application Perspective

This topic is an application-oriented reference for multimedia database technolo-
gies. It reports the progress in the area that has led to a range of innovative
applications and served in a number of different user scenarios, such as:

￿ Search and retrieval of multimedia objects , where a user from the local or remote

terminals submits a query (in the form of audio, photos, texture, video files) to a
search engine containing a large collection of multimedia repositories.

￿ Mobile social network , where users exchanges multimedia information among

friends or people located in a certain area, possibly enriched with recommen-
dation data related to points of interest. The users use the multi-touch screen
to focus on the visual intent, and the visual recognition system recommends
interesting restaurants, which may be incorporated with contextual information,
such as geo-location, for location-based recommendations.

￿ Landmark recognition , whereby a user captures the landmarks/scenes that he

or she is interested in, and obtains relevant information about the captured
landmarks/scenes.

￿ Movie/sports-video-on demand , where a single long duration video is retrieved

from a pool of servers, each of which retrieves disjoint portions of a movie
in a particular sequence to form one complete continuous stream for the
user. The topic presents the indexing, retrieval, and classification methods for
movie databases, which can be potentially adopted for the movie-on-demand
application.

￿ Interactive television news service , where a user employs a news headline, which

represents the subset of story, to retrieve full new stories, and find further relevant
stories in newscasts.

￿ Digital forensic investigation support , where an investigator utilizes digital image

(i.e., cartridge-based case image) taken at the crime scenes to query a forensic
reference database to search for digital evidence, which can allow conclusions to
be drawn for further physical investigation.

￿ Virtual reality motion/dance training , where a student dance is automatically

captured and used as a query to retrieve similar teacher motion data in a database
for the student's assessment; or where a user submits a short query motion clip,
and the task is to retrieve all clips in the database containing parts or aspects
similar to the query for data-driven computer animations.