2.2 MPEG-7 and Multimedia Database Systems

< Day Day Up >

Exhibit 2.3 shows a possible use of MPEG-7 in a distributed multimedia database system. Starting from the annotation process (automatic tools or manual annotation) the MPEG-7 descriptions are generated and stored for further access. The indexing process can be supported by stored information. For example, to annotate a new soccer video, we may rely on information of the players already stored in the database to avoid multiple annotation of the same player. Note that multimedia metadata may be available before the video is available. For example, the information on location and the team players is in general available before the soccer game takes place.

Exhibit 2.3: Usage of MPEG-7 in a multimedia database system.

click to expand

The query scenario distinguishes two cases: pull and push. If we consider a pull scenario, a user submits queries to the multimedia database and will receive a set of descriptions matching the query for browsing (for inspecting the description, for manipulating it, for retrieving the described content, etc.). Alternatively, if a streaming server is available, the user receives the media, together with the descriptions.

In a push scenario, a filter (e.g., an intelligent agent) will select descriptions from those available and perform the programmed actions afterward (e.g., switching a broadcast channel or recording the described stream). For this scenario, it is advantageous to encode and index the video in binary format for a fast-randomized access. In both scenarios, all the modules may handle descriptions coded in MPEG-7 formats (either textual or binary).

A pull scenario has been described in the introduction (Chapter 1). A user records 10 seconds of a radio song (query song) and sends it to a mobile service that relies on a multimedia database for recognition of the title and singer. The multimedia database extracts the AudioSignature of the song and generates an MPEG-7 description of the query song. This description is compared to the signatures stored in the audio database, and the most similar signature is retained. The system retrieves the title and singer information for this result signature and returns them to the user.

A typical push scenario is related to digital television broadcast enhanced with MPEG-7 information. Assume that a television viewer has expressed in MPEG-7 his or her preferences about program content (User-Preferences DS). For instance, they've requested that only sport news of a television news program be presented. In this case, the MPEG-7 descriptors in the broadcasted stream are filtered according to the MPEG-7 User-Preferences descriptions available, and only the sport news program is shown.

From the MPEG-7 metadata available with the broadcast stream, the user might identify additional material that is specifically of interest to them. For example, imagine a user is watching "Never Say Never Again." That user may decide to seek further information regarding that film (e.g., other films with Klaus Maria Brandauer) or simply other related data available on the Internet (e.g., price of the film's soundtrack on CD).

Many technical issues have to be addressed here. Just to mention a few: How to retrieve the features? In what format to store the MPEG-7 document? What query language to use? How to reference the media described? These issues are explored in Chapters 4 and 5.

< Day Day Up >