• Ingen resultater fundet

Why This Workshop ?

N/A
N/A
Info
Hent
Protected

Academic year: 2023

Del "Why This Workshop ?"

Copied!
10
0
0

Indlæser.... (se fuldtekst nu)

Hele teksten

(1)

Workshop 6:

MPEG-7 Audio:

What is it about ?

Chair: Jürgen Herre

Fraunhofer Institut for Integrated Circuits (FhG-IIS) Erlangen, Germany

(2)

Dr. Jürgen Herre, hrr@iis.fhg.de Seite 2

Why This Workshop ?

• MPEG Audio standards: A success story

• Past workshops about MPEG-4 Audio (Version 1 @ 106th AES, Munich;

Version 2 @ 108th AES, Paris)

• Next member of MPEG standards family:

MPEG-7, to be published in 10/2001

• It's time for an update ...

(3)

MPEG-1 (1992) MPEG-2 (1994)

MPEG-2 AAC (1997) MPEG-4 (1999+)

MPEG-7 (2001)

MPEG-7 or: “Is there an MPEG life after source coding ?”

“Moving Pictures Expert Group”, ISO/IEC JTC1/SC29/WG11

• First generic audio coding standard, Layers 1-3, (DAB, Worldspace, DVB, Internet Audio/”MP3”)

• Extending MPEG-1 coders towards lower sampling rates & multi-channel...

• More powerful mono ... multi-channel coding

• New functionalities (scalability, object oriented representation, interactivity ...)

• “Multimedia Content Description Interface”

Metadata standard (not compression!) MPEG

(4)

Dr. Jürgen Herre, hrr@iis.fhg.de Seite 4

Drowning in data ...

How to find ?

Vision

Explosive Growth of Available A/V Data

• Rapid growth of available A/V material on world wide scale

• Huge A/V databases → WWW/Internet !

• How is efficient search for multimedia content possible ?

• Currently: Efficient text-based search is provided by well-known search engines (e.g. Yahoo, Lycos, ...)

• “Audiovisual data should be just as

‘searchable’ as text!”

• Enable intelligent navigation & search

(5)

The Concept of Content Description (“Metadata”)

• Supplement A/V data with Content Description (“data about data”)

• linked to actual A/V data, but not necessarily on same location/media

• independent of format of actual A/V data

• characterize

• enable search / filtering

• enable navigation

• ...

→ enable efficient content handling

→ see also Workshop W-14 (Tue) ! Idea

A description is ...

Descriptions ...

(6)

Dr. Jürgen Herre, hrr@iis.fhg.de Seite 6

How to Generate Descriptions

Types of extraction/creation:

• Signal-based attributes - low semantic level (level, pitch, color, shape, ...)

• Higher level attributes (title, composer,

scene description including persons / objects

• Frequently descriptions are created as a by- product of the production process

(e.g. storyboard for a movie production) Automatic

(Extraction) Manually

Note

(7)

What are the issues ?

Areas of Work in the “Metadata” Context

Advancement of extraction methods

• automatic / semi-automatic

• Analysis of complex content

• Semantically meaningful

• Efficient

Standardization of description formats

• Enables interoperability between

metadata databases and applications on world wide scale

• Some standards already on the way (SMPTE, EBU, Dublin Core, MPEG-7, ...)

(8)

Dr. Jürgen Herre, hrr@iis.fhg.de Seite 8

The MPEG-7 Standard

• Standardization of “description language”:

• Extensible by means of Description Definition Language (XML-based, describes/defines

MPEG-7 syntax)

• Provides elaborated concepts for describing – Generic properties of A/V data

– Visual data – Audio data

Description production

Standard description Description consumption Boundaries of the

MPEG-7 standard

(9)

Questions To Be Answered ...

• What is MPEG-7 Audio about ?

• What are the basic concepts ?

• What novel functionalities can MPEG-7 Audio give you ?

• What are the applications ?

• “Could we see some demos ?”

(10)

Dr. Jürgen Herre, hrr@iis.fhg.de Seite 10

The Presentations ...

Adam Lindsay

University of Lancaster, UK Youngmoo Kim

MIT Media Labs,USA Philip Garner

Canon Research Center Europe, UK Geoffroy Peeters

IRCAM, France Michael Casey MERL, USA Jürgen Herre

Fraunhofer IIS, Germany Introduction of

MPEG-7 / MPEG-7 Audio Melody Description

Spoken Content

Musical Timbre Similarity

General Sound Recognition &

Similarity Tools

Robust Audio Matching

Referencer

RELATEREDE DOKUMENTER

Note: This Figure is based on bi-annual data provided by Eurostat for consumption band DC: 2,500-5,000 kWh (household electricity consumption) for Albania (AL), Bosnia and

This approach uses Generalized Search Tree (GiST), a data structure that provides all the basic search tree logic required by a DBMS, unifying different structures like B + -trees

Zeeker Search Engine has many promising features and we have no doubt that it can be made even better by correcting the errors found and adding some of the future features

Considering the use of OpenHub is known although not used by all contributors, the user data might not be accurate and a project with 100 users are actually known and well used by

Textual similarity measures are heavily used in text retrieval systems such as search engines or tools used for detecting plagiarism or copyright infringement of texts where a

Interface - EBSCOhost Search Screen - Advanced Search Database - CINAHL with Full Text. S25 behavioural pain score

• the schema‑based work lows and search interfaces − complex data sets visualization, navigation across hierarchical directory structures, adaptive queries and building

Source: Calculations by Danish Technological Institute based on a search in global patent databases via PatSnap.. 28.398 patent