Databases Reference
In-Depth Information
meanings on sensitive recognition, such as “human senses,” “feelings,” “sensitivity,”
“psychological reaction” and “physiological reaction”
17)
.
In the field of database systems, the concept of Kansei is related to data definition
and data retrieval with Kansei information for multimedia data, such as images,
music and video. The important subject is to retrieve images and music dynamically
according to the user's impression given as Kansei information. We review some
multimedia systems which manipulate Kansei information for defining and
retrieving multimedia data.
As discussed in
17)
, the field of Kansei was originally introduced as the word
“aesthetics” by Baumgrarten in 1750. The aesthetica of Baumgrarten had been
established and succeeded by Kant with his ideological aesthetics. In the research
field of multimedia database systems, it is becoming important to deal with Kansei
information for human beings for defining and extracting media data according
to impressions and senses of individual users.
The conceptual overview of a Kansei database system is shown in Fig. In the
Kansei database system, the essential functions for dealing with Kansei in database
systems can be summerized as follows:
(1) Defining Kansei information to media data (metadata definition for media
data).
(2) Defining Kansei information for user's requests (metadata definition for
user's requests (user's keywords) with Kansei information).
(3) Computing semantic correlations between Kansei information of media data
and a user's request (media data retrieval subsystem with a correlation computation
mechanism).
(4) Adapting retrieval results according individual variation and improving
accuracy of the retrieval results by applying a learning mechanism to metadata
(learning mechanism for metadata).
There are several research projects to realize these functions. In the design of
the Kansei information for media data, the important issues are how to define and
represent the metadata of media data and how to extract media data dynamically
according to the user's impression and the data contents.
Creation and manipulation methods of metadata for media data have been
summerized in
18,
19,
2)
. Furthermore, some research projects for Kansei information
have been established in academic fields
17)
. As one of Kansei research projects,
“modeling the evaluation structure of Kansei” has started in 1997
17)
. Multimedia
database subjects related to kansei information retrieval are promoted in this
project.
There are many approaches to media data retrieval. Two major approaches are
direct retrieval using partial pattern matching and indirect retrieval using abstract
information of images.
Several multimedia database systems for Kansei information retrieval have
been proposed. The pictorial information server systems, named TRADEMARK
and Electrical Art Gallery Art Museum, have been proposed to perform picture
retrieval using query-by-visual-example and query-by-subjective-descriptions
3)
.
