d3 6 3 - acoustics research institute · 2012-12-06 · 3.7 signal enhancement and signal...

HARMONICAHarmonised Access to Music and Music Information in Libraries

Libraries Project: PROLIB/HARMONICA 10453Commission of the European Communities

LIBRARIES PROGRAMME

Remote Access and Transfer of Audio Recordings

Deliverable Number: D3.6.3

Version: 1.0

Date: 30. 11. 1999

Authors: Werner A. Deutsch, Siegbert Herla , Werner Kriechbaum

Confidentiality: Public

Status: Draft

This document consists of <35> pages plus this cover

PAGE DOCUMENT VERSION1 D3.6.3 Access & Transfer 1.0

CONTENTS

1. INTRODUCTION ..................................................................................................1

1.1 Purpose and Scope ..........................................................................................1

1.2 Applicability .......................................................................................................1

1.3 Acronyms and Abbrevations .............................................................................2

2. ANALOGUE AUDIO.............................................................................................4

2.1 Analogue Tape-Formats ...................................................................................4

2.2 Digital Audio Formats........................................................................................4

2.3 DVD - A Breakthrough in Digital Audio? ...........................................................5

3. DIGITISATION .....................................................................................................6

3.1 General .............................................................................................................6

3.2 Digital Audio Workstations (DAW) ....................................................................73.2.1 Analogue to Digital Conversion – Dynamic Range 73.2.2 Sample Rates and the 24 – 20 –16 bit Story 73.2.3 Linear PCM Audio 16-bit is not obsolete! 8

3.3 New Digital Audio Formats................................................................................93.3.1 Archive File Formats 93.3.2 Broadcast WAVE Format (BWF) 93.3.3 ‘Unique’ Source Identifier (USID) 11

3.4 BWF - A Music Library Audio Format?............................................................12

3.5 De-Facto/Industry Digital Audio Compression Standards ...............................12

3.6 Digitisation Procedure Quality Control ............................................................13

3.7 Signal Enhancement and Signal Restoration..................................................13

4. AUDIO SEGMENTATION AND CONTENT DESCRIPTION..............................15

4.1 Query by Audio Content (QBAC). ...................................................................154.1.1 Sound File Segmentation 154.1.2 Links Refering to Segments 16

4.2 Content Description ........................................................................................19

4.3 Visualisation of Music Signals.........................................................................21

4.4 Future Development of Content Driven Approaches: Audio DescriptionSchemes ................................................................................................................22

4.4.1 The MPEG-7 Audio Descriptor Scheme (tentative) 23

5. A MODULAR ARCHIVE MODEL.......................................................................25


5.1 Acquisition of documents and information.......................................................25

5.2 Archival Storage..............................................................................................26

5.3 Data Management...........................................................................................26

5.4 Administration .................................................................................................27

5.5 Access ............................................................................................................27

6. AUDIO-NETWORKING......................................................................................28

6.1 The Role of Libraries and Archives on the Internet .........................................286.1.1 General 286.1.2 A Glimps on Copyrights 29

6.2 Library and Archive Services on the Internet ..................................................296.2.1 Connectivity 296.2.2 Delivery Model 30

6.3 MP3 – An Evolving Digital Music Delivery Sector ...........................................31

6.4 Frequently Used Bit Rates ..............................................................................32

7. APPENDIX .........................................................................................................33

7.1 A Sample List of Audio Players.......................................................................33


1. Introduction

1.1 Purpose and ScopeThe purpose of this document is to provide an overview and references toinformation on transfer of analogue audio recordings into digital formats(digitisation), local and remote storage of data as well as access and re-trieval in a library or archive environment. A description of current state ofthe art data acquisition workstations, quality control and archive referencemodels is given. New developments in network capabilities providing betterservice to selected network traffic over various technologies (QoS: Qualityof Service networking) are in discussion.

• Analogue Audio: formats still working

• Digital Audio Workstations (DAW): data acquisition, local storage,segmentation, technical metadata generation, access issues

• Archive Reference Model: ingest station, archival storage, data man-agement, administration, access

• Local area network (LAN) configuration: Gigabit, ATM, QoS, tailoredservices, special requirements of streaming media

• Internet connectivity: FTP, MP3, Real Audio

1.2 Applicability

The issues described in this document may be applicable to any soundstorage, archive or collection. They are applicable to organisations with theresponsibility of providing information on a temporary basis as well as forthe long term. When taking the rapid pace of technology changes or possi-ble changes in a Designated Community into consideration, there is thelikelihood that facilities thought to be holding information on a temporarybasis will in fact find that some or a lot of their holdings will need the samekind of attention as that given by permanent archives.

The deliverable is not intended to offer a de facto standard for digital sounddocument engineering, but rather as help for digitisation, editing, tagging,indexing and storage. It is aimed towards library and archive institutionsthat already have or are building up the equipment and expertise to digitisesound documents in-house. It addresses the more standard formats ofsound. If it is planned to digitise primarily historic or fragile and rare sounddocuments or materials in non-standard formats and sizes, it might be con-sidered outsourcing the digitisation of these materials to specially equippedlaboratories or institutions1.

1 See D3.6.4 In-house pros/cons. Outsource pros/cons.


1.3 Acronyms and Abbrevations

A/D - Analogue/DigitalAES - Audio Engineering SocietyAIC - Archival Information CollectionAIP - Archival Information PackageAIU - Archival Information UnitASCII - - American Standard Code for Information InterchangeBEXT - Broadcast Extension Chunk (BWF)BWF - Broadcast WAVE file formatCAD - Computer-Automated DesignCAR - Computer-Aided RadioCCSDS - Consultative Committee for Space Data SystemsCD-ROM - Compact Disk - Read Only MemoryCIP - Catalog Inter-operability ProtocolCRC - Cyclical Redundancy CheckD/A - Digital/AnalogDAPA - Digital Audio Production and Archiving (EBU working group)dB FS - Decibel (relative to full scale value)dB m - Decibel (relative to 1 mW)dB r - Decibel (relative to an absolute reference)dB SPL - Decibel (relative to 20 ìPa)dB u - Decibel (relative to 0.7746 V) equiv. to dB m at 600 �dB v - Decibel (relative to 1 V)dB - Decibel (1/10 Bel)DBMS - Data Base Management SystemDDL - Data Description LanguageDED - Data Entity DictionaryDFAS - Distributed Finding Aid ServerDIP - Dissemination Information PackageDLI - Digital Libraries InitiativeDR - Dynamic RangeDSD - Direct Stream Digital (1-bit delta sigma technology: SACD)DTD - Document Type DefinitionDVD - Digital Versatile DiskDVD-Audio - DVD working group specification audioEAD - Encoded Archival DescriptionEBCDIC - Extended Binary Coded Decimal Interchange CodeEBU - European Broadcasting UnionERL - Electronic Reference LibraryFITS - Flexible Image Transfer SystemGIF - Graphics Interchange FormatHDCD - High Definition Compatible Digital format (Pacific Microsonics)HFMS - Hierarchical File Management SystemHFS - Hierarchical File ServerHTML - Hypertext Markup LanguageICS - Interoperable Catalogue SystemIEEE - Institute of Electrical and Electronic EngineersIMS - Information Management SystemISBN - International Standard Book Number


ISO - Organisation for International StandardisationLSB - Least Significant BitMPEG - Motion Picture Expert GroupMPEG 1 - Coding for Moving Pictures and Associated Audio for Digital

Storage Media up to 1.5 MBit/sMPEG 2 - Generic Coding for Moving Pictures and Associated Audio

(multichannel audio)MPEG 1 Audio Layer 1 Audio Coding Scheme (compression ratio: 1:4)MPEG 1 Audio Layer 2 Audio Coding Scheme (compatible to Layer 1)MPEG 1 Audio Layer 3 Audio Coding Scheme (compatible to Layer 1 & 2)MP3 - syn. MPEG-1 Audio Layer 3 (compression ratio 1:10...12)NARA - National Archives and Records AdministrationNASA - National Aeronautics and Space AdministrationNSF - National Science FoundationOAIS - Open Archival Information SystemOCR - Optical Character RecognitionODL - Object Description LanguageODLS - Oxford Digital Library ServicesOPAC - On-Line Public Access CataloguePCI - Periodicals Contents IndexPDI - Preservation Description InformationPDMP - Project Data Management PlanQBAC - Query By Audio ContentQBIC - Query By Image ContentQoS - Quality of the ServiceRIFF - Resource Interchange File FormatRLG - Research Libraries GroupSACD - Super Audio CD (Sony, Philips format, uses DSD)SGML - Standard Generalised Markup LanguageSIP - Submission Information PackageSuper CD - DVD-based audio formats, superior to audio CDs.TEI - Text Encoding InitiativeUML - Unified Modelling LanguageUNICODE - Universal CodeWAV - Windows Wave FileWWW - World-Wide Web


2. Analogue Audio

The majority of historic sound documents and many recent digital remakesare still based on analogue audio formats. Analogue audio technology hasmatured to an excellent high level of sound quality from the early begin-nings of non-linear Thomas Alva Edison´s horn recordings to the almostlinear transmission chain of today. Generations of audio engineers haveoptimised the linear transfer function of tape recorders, amplifiers and disccutting machines in order to produce noiseless and distortion free record-ings. It was left to our generation to re-introduce nonlinearity when percep-tive coders as MP3 are applied in broadcasting and network environments.Perceptive coders perform lossy coding omitting the possibility to recon-struct the original signal. Nevertheless, MP3 sounds acceptable, allows lowbandwidth connections and has yet developed to one of the most frequentlyused audio formats for audio transmissions over the Internet.

The European Broadcasting Union (EBU) lists considerable less analogueaudio tape formats as digital ones being referenced in the broadcastingarea. This (4/1997) list is remarkably conservative and several additionaldigital formats have been introduced in the meantime (see also Section 3 ofthis document):

2.1 Analogue Tape-FormatsA01 6.3 mm analogue audio Full trackA02 6.3 mm analogue audio 2 channelA08 12.5 mm analogue audio 8 channelA16 25.4 mm analogue audio 16 channelA32 25.4 mm analogue audio 32 channelAS2 6.3 mm analogue audio 2 channel stereoAT2 6.3 mm analogue audio 2 channel stereo & TCCCA Compact Cassette audio

2.2 Digital Audio Formats

CDA Compact Disc AudioD24 25.4 mm digital audio DASH 24 trackD32 25.4 mm digital audio PD 32 channelD48 25.4 mm digital audio DASH 48 trackDA2 DAT format digital audio 2 channelDAT DAT format digital audio StereoDD2 6.3 mm digital audio DASH 2 channelDP2 6.3 mm digital audio PD 2 channel3.5" data diskette - FD55.25" data diskette - FD88" data diskette - H8AHi-8 digital audio 8 channelMO disk 600 MBytes capacityM12 MO disk 1 200 Mbytes capacity


M13 MO disk 1 300 Mbytes capacityNAB NAB audio cartridge -S16A-DAT digital audio 8 channel

2.3 DVD - A Breakthrough in Digital Audio?

Although people have always been full of expectation for improved soundquality available from higher sample rates and greater resolution, technicallimitations have made it impractical to implement a new standard until now.The arrival of DVD — the Digital Versatile Disk — has the potential oftransforming the possibility for improved digital sound. A DVD offers overseven times the storage capacity of a CD. This additional capacity allowsfor an improvement in the resolution of the audio signal as well as in sam-ple rates. The DVD specification allows for – and manufacturers of audioequipment and programme material (originally a group of 10 members washolding more than 4000 DVD related patents) use - many different resolu-tion levels; but the dominant standard for high quality audio samples at 96kHz to a resolution of 24 bits. This translates to 16,777,216 (2^24) differentpossible amplitude levels at a theoretic dynamic range of 144 dB. Com-bined with a more than doubled sampling frequency, DVD audio offers over500 times the resolution available from CD. DVD audio provides for the firsttime in audio engineering the technical potential to produce and distributemusic at a sound quality considerably better than human listeners can hear.

To probe further:http://www.music.mcgill.ca/~martin/bibliography/digital.html

For a survey on historic sound formats and preservation issues see Har-monica Deliverable D3.1 “Analogue Documents, Carriers and Formats”http://www.kfs.oeaw.ac.at/harm/D3x1.htm


3. Digitisation

3.1 GeneralAlthough many digitisation projects will be done for purposes of increasedaccess to sound archives and library collections primarily, preservation isoften a natural by-product. Digitisation should therefore be performed with a"preservation mindset." This mindset implies2:

• Performing analogue to digital (A/D) conversion and digital to analogue(D/A) conversion at the highest sample rate appropriate to the natureand the informative content of the originals

• Performing analogue to digital conversion at an appropriate dynamicrange and sound quality to avoid re-doing the transfer and re-handling ofthe originals in the future - digitise once only

• Creating and storing a linear coded master sound file that can be usedto produce derivative and compressed sound files in order to serve a va-riety of current and future user needs (i.e. data reduced and perceptualcoded copies for browsing and Internet access)

• Using system components that are non-proprietary

• Using sound file formats, editing systems and data compression tech-niques that conform with industry standards

• Creating backup copies of all files on a stable medium

• Creating meaningful metadata for sound files and associated documentsincluding cataloguing issues (if appropriate)

• Monitoring and recopying data if necessary

• Outlining a migration strategy for transferring data across generations ofarchive and access technology (plan for obsolescence of current hard-and software technology)

• Anticipating and planning for future usage and technological develop-ments

This document occasionally suggests the minimum hardware standards aswell as high end professional audio equipment but libraries and archivesshould not just "do the minimum." Analogue to digital conversion at a highersample rate and resolution rather than to the minimum required is encour-aged. One plausible argument in favour of adapting a higher technical

2 See: COLORADO DIGITIZATION PROJECT: GENERAL GUIDELINES FOR SCANNINGwhich correspondingly apply for sound documents.http://coloradodigital.coalliance.org/cdp.html


standard from the beginning is given by the development of costs during adigitisation project. Cost relation for equipment and staff usually range 1:1at the start of a digitisation project and drops to 1:4 after a 4 year period.

3.2 Digital Audio Workstations (DAW)

3.2.1 Analogue to Digital Conversion – Dynamic Range

The main function of a digital audio workstation is performing a high qualityanalogue to digital conversion of a continuous audio stream generated froman analogue signal source as well as converting the digital audio streamback to an analogue signal. The quality of the conversion is determined bythe resolution available from the analogue to digital converter (A/D) anddigital to analogue converter (D/A). The resolution of an A/D – D/A con-verter system is given by:

The dynamic range of an A/D - D/A subsystem can conveniently be ex-pressed in dB. According to a resolution of 1*x(V), a n*bit system provides

i.e. � 96 dB at n=16 bits, � 108 dB at n=18 bits, � 120 dB at n=20 bits and �144 dB at n=24 bits.

3.2.2 Sample Rates and the 24 – 20 –16 bit Story

Sample rates in professional audio and video environment range from 32kHz to 96 kHz and 192 kHz (DVD audio), 44.1 kHz and 48 kHz being usedmostl frequently. Other sample rates occasionally used are: 44.056 kHz,47.952 kHz, 64 kHz, 88.112 kHz, 88.2 kHz, 95.904 kHz and 176.4 kHz.

It is advisable for sound archives and collections to maintain audio work-stations capable to select sample rates ± 10% off the nominal values. Atleast one unit serving arbitrary sample rates between 5 kHz and 48 kHzshould be available in order to handle non-standard recordings from sev-eral sources; otherwise a sample rate converter as a separate functionalunit is necessary. An overview on popular sample rates and sound file for-mats is given in HARMONICA deliverable D3.2http://www.kfs.oeaw.ac.at/harm.

High end DAWs use accurate, discrete, multi-bit A/D converters. The A/Dconverters operate at a sample frequency of 192 or 176.4 kHz and employsophisticated digitally subtracted dither to produce both low noise and dis-tortion components below -120 dB FS, or less than one part per million.

resolutionsignal range V

bitsn=− ( )

2

L dBxn= ×20 2lg ( )


The 192/176.4 kHz signal is decimated to 96 or 88.2 kHz, 24 bits using op-timised filtering.

In spite of all technology progress CDs are still in 16 bit audio. Several so-lutions for handling both the 24 and 16 bit domain have been proposed. Inorder to reduce the 44.1 kHz, 24-bit signal to 16-bits while retaining many24-bit audio benefits, soft limiters are applied which allow the increase ofthe peak signal level up to 6 dB without overloading. The peaks are recon-structed when decoded increasing dynamic range by 6 dB. For undecodedplayback the units work as standard limiter. Some DAW units provide a lowlevel range extension which gradually increases the gain on low-level sig-nals (approx. starting at -45 dB FS) by 4 dB over a 20 dB range.

As one of several possibilities the final step in the reduction to 16-bits is toadd high-frequency weighted dither and round the signal to 16-bit precision.The dither can be applied to the frequency range of 16 kHz to 22.05 kHzleaving the noise floor flat below 16 kHz without influencing the psy-choacoustic relevant frequency range for the perception of tonal signals.

Psychoacoustically designed noise shaping filters are controlled by thespectral range of the time varying audio signal. Some audio systems intro-duce, as part of the final quantisation, a pseudo-random noise hidden codeas needed into the LSB of the audio data. The hidden code carries thedecimation filter selection and peak detection and low level range parame-ters. The hidden code is completely inaudible and is only inserted 2-5% ofthe time, effectively producing 16-bit undecoded playback resolution. Theresult is an industry standard 44.1 kHz, 16-bit recording which should becompatible with all CD replication equipment and consumer CD players.

Although DAW producers advertise their signal processing being compati-ble to CD-standards, careful verification of the format along CD-Red Bookspecification is necessary, which calls for linear PCM audio at a 16-bit wordlength and 44.1kHz sample rate. Special care has to be taken if sounddocuments from unknown sources are not accompanied by the appropriatedigitisation side information. Chaining of different noise shaping or ditheringconcepts can produce distortions quite above the hearing threshold, al-though each of them is inaudible when listening to them alone.

3.2.3 Linear PCM Audio 16-bit is not obsolete!

Audio technology has to be seen transitional today. With both DVD-Audioand Super Audio CD on the horizon, libraries and archives will face newchallenges before classic digital audio formats have been acquired suffi-ciently. A psychological turning point will probably come when the CD is nolonger seen as the best available sound quality format by the public. Justas the CD in comparison to the LP and the cassette on its way up, the newformats will gain consumer acceptance rapidly. Nevertheless, CD format:linear PCM audio at a 16-bit word length and 44.1kHz sample rate is notobsolete. It will remain the only viable consumer digital delivery formatbuilding the mainstream in consumer audio electronics for quite a while. Anestimation of millions of CD-recorders to be sold per year represent a lot of


hardware for the survival of the format. Recordable DVD is still in discus-sion on different formats (and size: 12 cm, 8 cm?); the projection that oneDVD can possibly not be run on two different laptops is confusing thesalespeople as well as the users. DVD as Digital Video Disk will even de-pend on the manufacturers choice to make it playable in more than onegeographical region.

3.3 New Digital Audio Formats

3.3.1 Archive File Formats

In the early 90s, computer-aided radio (CAR) systems became digital audioislands in broadcasting houses. These CAR systems used propriety filesystems. The music and radio programme exchange between different is-lands took a lot of processing time just for file conversion. Out of a varietyof sound file formats two evolved as a de facto standard: AIFF used in theMAC/UNIX world and RIFF/WAVE in the PC domain. This scenario wasfound by an EBU project group P/DAPA (Digital Audio Production and Ar-chiving) when negotiating with industry in order to propose a common fileformat for linear audio quality serving the AES/EBU hardware interfacestandard. In order to generate and process descriptive information con-veniently, metadata should also be included in the file format. The groupdecided to select the widespread RIFF/WAVE format as a proposal for astandard. One major advantage of the WAVE file format can be used:WAVE files are worldwide native files on all PC platforms and each PC isable to play and edit them. WAVE files are also used for audio data importand export on several other computer platforms. In order to enable stan-dardised audio programme exchange the group developed the so-calledBroadcast Wave Format. The main issue consisted in the agreement on aspecial designed 'Broadcast Extension Chunk' (BEXT Chunk) for storage ofadditional metadata and descriptive information in sound files.

3.3.2 Broadcast WAVE Format (BWF)

As libraries and archives receive music documents from quite differentsources and their users occasionally are located in a broadcasting envi-ronment, in future many sound files stored in BWF format may appear. Al-though libraries could consider BWF as a useful lib-rary standard, it seemsquestionable to convert all digital audio holdings into BWF as long as audiofile transfer with broadcasters is not needed often. BWF stands for a com-prehensive method to include metadata and links to additional descriptiveinformation in sound files but alternative solutions taking advantage fromthe possibility to define user chunks in the RIFF/WAVE format can be ex-pected to arise. It should be emphasised that sound data of standard linearcoded WAVE files remain playable with any wave-player available whetheror not additional chunks are included; whereas the content of user chunksneeds to be managed by special application software components.


Fig. 1: the Broadcast WAVE Format (from EBU Technical document 3285).

The Broadcast Wave Format (BWF) has been defined in EBU standardN22-1997. The full specification of BWF, a description of the BroadcastExtension Chunk and basic information on Microsoft RIFF format is given inEBU standard document Tech. 3285. BWF incorporates ISO/MPEG-2Layer II which is intended to be used as browse quality in sound archives.A description of MPEG support in BWF is given in Supplement 1 of Tech.3285. Further information on BWF as well as the documentation describedcan be obtained from the EBU (http://www.ebu.ch/pmc_bwf_ug.html)and from Swedish Radio Corporation (http://www.sr.se/rd/bwf/). In addi-tion to the BWF specification, the project group (P/DAPA) published thefollowing recommendation R85 for programme exchange of audio datafiles:

• · Sample rate: 48 kHz• · Resolution: minimum 16 bit / linear• · Alignment level: according to EBU R68 (headroom of 9 dB)• · Preemphasis: none• · Channel formats: mono, 2 channel stereo• · Signal formats: multi channel >> MPEG, linear PCM• · BEXT chunk: transfer ahead of the audio data

BWF does not support all types of RIFF chunks; nevertheless it is compati-ble to ISO OSI layer model for information interchange. BWF files can beused independently from the transport layer for data exchange in real time


and file transfer over networks as well as for signal storage on data carrierssuch as disks or tapes. One single BWF file is capable to hold about 4Gbytes of data, corresponding approximately up to 6 hours of linear stereosound signal (16 bit / 48 kHz) or 4 hours (24 bit / 48 kHz). This size of datavolume is sufficient for the storage and reproduction of almost all analoguesound carrier volumes commonly used.

The following BWF file structure is currently supported:

The BEXT Chunk contains general metadata such as title, originator,archive number etc.

A Coding History (part of BEXT chunk) describes the transmissionchain of the current sound signal, providing information such as ofsound carrier material, recording and playback equipment, analogue-to-digital-converter and digital I/O interface card of the PC.

The Format Chunk is used to specify format information as linear PCM,stereo, sample rate and resolution (16...24 bits).

The Quality Chunk contains information obtained from the digitisationprocedure, such as a protocol of defects in the analogue recording andtransmission chain, tape drop-outs, clicks, thumps, hiss, print-throughand additional notes (in preparation).

The Cue Sheet Chunk provides cue points, tags and segment data asoffset, start time and duration of a specific content in the file (in prepara-tion; see also Audio Segmentation and Content Description).

The Wave Chunk contains the audio samples of the digitised soundsignal.

3.3.3 ‘Unique’ Source Identifier (USID)In order to idetify BWF source sound files an unique identifier has beenproposed which serves as a prime link between sound files and associa-ted data in a database system. Applications can use the identifier insteadof the file name for reference.

The EBU proposed the following structure:OriginatorReferenceThe <OriginatorReference> field is as a sequence of 32 ASCII charac-ters (not a string) provided in the BWF to contain an unique identifier ofthe file. The organisation originating the BWF file is responsible for theallocation of the USID.Country code: (2 characters) is based on the ISO 3166 standardCompany code: (3 characters) is based on the EBU Technical informa-tion I30-1996.Serial number: (12 characters extracted) This should identify the ma-chine’s type and serial number.


OriginationTime (6 characters, HHMMSS) This should be sufficient toidentify a particular recording in a human-useful form in conjunction withother sources of information, formal and informal.Random Number (9 characters 0-9) Generated locally by the recorderusing some reasonably random algorithm. This number serves to sepa-rate files made at the same time, such as stereo channels, or trackswithin multitrack recordings.

Example of an USIDGenerated by a Tascam DA88, S/N 396FG347A, operated by Radiote-levisione Italiana, at time: 12:53:24

UDI format: CCOOOSSSSSSSSSSSSHHMMSSRRRRRRRRRUDI Example: ITRAIDA88396FG347125324098748726

3.4 BWF - A Music Library Audio Format?

As has been pointed out, BWF is an advanced real life example for com-bining sound and (technical) metadata in a comprehensive file format forbroadcasting storage and retrieval applications as well as for programmeexchange between different partners. One of the outmost advantages ofthe approach transporting all necessary metadata within the sound filesthemselves is easy management of sound files and metadata in the LANand across different hard- and software platforms. For the archive or librarystarting from scratch, a medium size digitisation project can be evolved onlearning by doing and just by implementing a minimum standard as BWFrequests. For up to several ten thousands of sound files of a homogenouscollection, general purpose computer systems and file archive servers pro-vide file management and browsing tools convenient for use.

Larger collections will not succeed without additional data base manage-ment and a specified file system structure. Links from (or to) an existingcatalogue have to be updated at regular intervals and sound and metadatashould be “frozen”. In any case, before digitising, libraries and archivesshould think very carefully about implementing appropriate file naming con-ventions for sound files and associated documents along content relatedclassification or indexing schemes in order to support effective data re-trieval later on.

The current BWF structure is certainly superior when the percentage ofmetadata associated to sound is small in comparison to the data size of thesound itself. It is applicable when search and browsing is primarily done onthe basis of listening to sound data and not by cruising through large vol-umes of content related metadata. One very useful extension of the BWFconcept could be providing appropriate links to recently evolving contentdescription standards as MPEG-7 promises to become.

3.5 De-Facto/Industry Digital Audio Compression StandardsAlthough sound archives and music libraries should not even think about toselect any perceptual (lossy) coded sound format as an internal archive


standard, they certainly will have to deal quite a lot with manyl sounddocuments encoded in different industry audio compression schemes. Ar-chives and libraries should therefore be prepared to read all formats onwhich relevant content will be stored as well as they could provide an usersevice to convert nonstandard formats into generally readable ones. Todayaudio formats are dictated by the music contents the consumer appreciatesrather than by technical quality criteria which would support technically op-timized solutions at reasonable costs: A list of audio coders and links touseful pages on this issue always is incomplete at that point in time it wouldbe written (for further information see Section 6 of this deliverable andamong others, http://www.research.att.com/~gjm/imusic/links.html).

3.6 Digitisation Procedure Quality Control

Real time sound analysis of the analogue and digitized audio stream isperformed during digitisation in order to control the analogue to digital con-version process. For this reason, digital audio workstations provide severaluseful features such as:

• · automatic detection of start and end position of the audio signal• · automatic detection of pauses during the recording• · automatic detection of the noise floor level• · automatic detection of clicks and impulsive distortions• · automatic detection of analogue media drop-outs• · average value of signal to noise ratio of the audio signal• · average value of frequency bandwidth of the signal• · average value of stereo correlation• · average value of level dynamics

Signal parameters extracted by means of digital signal processing as listedabove are used to control the sound quality of the digitised waveform. Anyerrors occuring during digitisation of analogue recordings should be de-tected on the fly. Standard quality criteria, obtained from long term statisticsare matched against actual values measured. A transfer (quality) protocol isusually added to the technical metadata set.

3.7 Signal Enhancement and Signal Restoration

Although Quality Control may report signal degradations which could be re-paired by means of digital signal enhancement algorithms, a strictly lineararchive copy should be produced in any case. This advice is justified be-cause any signal enhancement to be considered can be performed ondigital copies of the linear archive document with no loss of generality. Asdigital copies are 100% exact replica of the original no information losstakes place; whereas signal filtering or any other processing carried out al-ready at the time of digitisation would inevitably introduce nonlinear transfer


functions which frequently cannot be inverted later on. As a general rulenonlinear manipulations of the audio signal, such as lossy data reduction,compression algorithms, filtering and signal restoration, should be carriedout at the end of the transmission chain only and never on archive copies.

Some manufacturers of digital audio workstations and audio softwarepackages provide typical audio signal processing algorithms, working eitherin the time or frequency domain, such as:

• DeNoiser for the reduction of broadband noise (hiss)• RepairFilter for elimination of quasi-static noise (mains hum, hum

from dimmers, stereo pilot tones)• DeScratcher for elimination of scratches on vinyl record and shellac

recordings• DeClicker for automatic elimination of clicks• DeCrackler for automatic elimination of crackles• DeClipper for automatic elimination of digital clipping• DropOuter for automatic drop-out restoration• VPIs for remastering and sweetening

(parametric EQ, 1/3 octave EQ, linear phase EQ)

The list above can be completed by additional sound conditioning functionswhich belong to the standard equipment of a sound recording studio, usu-ally appreciated by the professional sound engineer; these are among oth-ers:Compressors, Limiter, Loudness Maximizer, De-esser, Free Shaper (redit-hering / noise shaping), stereobasis manipulator etc. For exact metering(level control) tools such as PhaseScopes (stereoscope) , FFT-Analyzers,1/3 octave Analysers, Real Time Spectrograms, MatrixScopes etc. areapplied.


4. Audio Segmentation and Content Description

4.1 Query by Audio Content (QBAC)3.

With the increasing acceptance of Digital Libraries and archives as storagesystems for music and multimedia data, efficient architectures for context-oriented search of non-textual data becomes a pressing need. Whereasboth, automatic shot detection and query by image content (QBIC), haveopened inroads to characterise and search for visual information, noequivalent methods exist for streaming audio data. Many audio data have acommon property which still images and video material do not have: It canbe expressed in two corresponding forms, either as a 'textual' representa-tion (music score, transcript) or as a realisation (sound recording).

Working with Streaming Media raises new issues related to collecting,storing, annotating, indexing, browsing, use of meta-data, and retrieval in-terfaces for libraries and audio/video archives. While in some cases, thereference linking of an entire audio file to a score or a text file might be suf-ficient just for "listen to", the correspondences of a search on the score ortext file require a fine-grained audio data segmentation. Once such a fine-grained linkage between textual representation (narrow transcription) andacoustic realisation is established, the textual representation can be usedto facilitate QBAC (Query By Audio Content). A language-aware searchengine locates the desired elements in the score or text. The result of thequery are segments of audio material (audio objects), identified by fine-grained links between score or text and the sound recording.

4.1.1 Sound File Segmentation

Sound segments (audio objects) are addressed sample by sample fromthe beginning of a recording. Usually, sample number offset and dura-tion of the segment is referenced. Segment identifiers are created byautomatic segmentation procedures or by programme supported man-ual editing. Cue in and cue out points, edit decision lists and play listscan be linked to segment identifiers and collected in sound file directorytables. The segment structure should not be limited to a single seg-mentation layer and relative addressing of segments should be sup-ported. In order to facilitate context related queries, overlapping seg-mentation should be implemented. Sound file directories may grow toconsiderable size (several thousands entries for each file) in the courseof cumulative segmentation work sessions. Furthermore, as one andthe same audio signal has to be segmented differently according tospecial user requirements, it seems appropriate to store the segmenta-

3 AC308 ACTS-DICEMAN: Distributed Internet Content Exchange using MPEG-7 andAgent Negotiations.


tion metadata in separate files which serve as input to an audio contentrelated database.

Archived sound data usually do not change anymore after digitising.What has to be updated in regular intervals are metadata and metadatalinks, the location of suitable cue in points, segment sequence proce-dures for rapid browsing, the creation of clips and several further ar-chive staff and user accessible functions. The concept to manage themetadata separated from the sound files enables fast and easy accessand virtual (non-destructive) processing and access of sound4.

Fig. 2: segmentation of sound files in multiple layers. Sound segmentaddresses, segment identifier, optional links and content description arestored in a sound file directory which is separated from the sound.

4.1.2 Links Refering to Segments

Several standard, industry-standard, and proprietary standards can beused for to express the links needed to refer to segments. However, sincein many cases neither the audio recording nor the description or otheraudio segment linked to it can or should be changed, a simple hyperlinkscheme as in HTML is not sufficient. Instead it is necessary to use so-called ‘independent’ hyper links, which are external to the files they link. Inaddition, these hyper links must be bi-directional (description to audio andvice versa) to allow both, applications like querying the text and playingback the speech, as well as playing the audio and switching e.g. to the dis-play of the score at an arbitrary point in time.

4 http://www.kfs.oeaw.ac.at/S_TOOLS/stxdoc/aprimer.htm

audio streams

sound file 1

sound file 2

sound file 3

sample 0

sample 0

sample 0

offset/duration/segment ID/optional links/content description..

links between segments of different sound files


Currently the most advanced linking mechanism available is the HyTimeilink [HyTime97]5. Similar functionality can be provided by other implemen-tations and will is likely to be provided by XML linking mechanisms. In thefollowing discussion SGML and HyTime will be used as an example. An in-dependent link consists of three components:

Anchors, which are regions or points in a text or audio document.Locators, which locate or address anchorsLinks, which link or connect locators.

LocatorLocator

Hyperlink

Anchor Anchor

bidirectional link

Figure 3: The canonical bi-directional indipendent link from HyTime.

The ISO/IEC HyTime standard offers locators to uniquely address elements(sub trees) within SGML documents. This mechanism assigns a list of inte-ger values to each node of an SGML tree. The list of integer values is the'road map' to get from the root of the SGML document (tree locator '1') tothe specific SGML element using several 'traffic rules' to generate the treelocator integer list. The rules are:

1. The 'journey' starts at the root element and adds one integer foreach horizontal level below on the way down the SGML tree.

2. The root element of the SGML tree has the tree locator '1'.3. Each integer stands for one horizontal level of the SGML tree.4. Each integer value is generated by counting the number of

nodes from left to right. Only the children of the node above aretaken into account.

5 [HyTime97] ISO/IEC JTC 1/SC18 WG8 N1920rev, “Information-Processing - Hypermedia/Time-based Structuring Language (HyTime) 2nd edition,” ed. Charles F. Goldfarb, Steven R.Newcomb, W. Eliot Kimber, Peter J. Newcomb. May 1997.


5. The left most node (left most child) of a node above is assignedthe integer value '1'.

Starting at the root element (tree locator '1') and taking all the above rulesto generate the tree locator into account, the nodes (elements) of the fol-lowing abstract tree will be addressed by the tree locators listed in the tablebelow.

Element Tree locatorA 1B 1 1C 1 2D 1 1 1E 1 1 2F 1 2 1G 1 2 2

Table 1: Tree locators for the abstract SGML tree:

In SGML, the DTD fragments needed for a link between audio and textwould be expressed in a form similar to :

...

<!ELEMENT audioloc - - (#PCDATA) >

<!ATTLIST audioloc

id ID #REQUIRED

HyTime NAME #FIXED "queryloc"

>

...

<!ELEMENT textloc - O (#PCDATA) >

<!ATTLIST textloc

HyTime NAME treeloc

id ID #REQUIRED

locsrc CDATA #IMPLIED

>

...

<!ATTLIST audio


HyTime NAME ilink

id ID #IMPLIED

linkends IDREFS #REQUIRED

>

...

Using above definitions, the link itself would be expressed in a form similarto :

...

<audiolink linkends="text audio">

<audioloc id="audio">>file=test.wav start=588 end=24703

unit=ms</audioloc>

<textloc id="text" locsrc=test.txt>1 1 2 1 1 1</textloc>

...

4.2 Content Description

In the previous section it has been shown that content description ofsound is closely related to narrow segmentation of sound files. Whereasthe automatic creation of video objects6 has already been successful byuse of semiautomatic analysis tools nothing comparable exists foraudio. The limitations experienced in the content description of stream-ing video equally apply – even to a larger extent – for audio data. Fur-ther development is needed for:

• segmentation tools requiring fully automatic, real-time analysis

• applications which may allow some convenient level of user guid-ance

• means to exploit the complementary skills of user and machine insolving complex analysis and interpretation problems

• tools that assist in the integration of the different solution frag-ments

• means to establish consistency of the final description, especiallyin corres-pondence to video sequences in case of multimediadocuments

• applications for the integration of audio content description in anexisting catalogue framework

Some open issues have been seriously addressed by the CUIDAD7

group, contributing to MPEG-7, building a bridge between low level

6 ACTS-MoMuSys (Mobile Multimedia Systems)7 CUIDAD is a European Working Group coordinated by Ircam - Centre Georges Pompidouin order to gather all institutions, industrials and users interested in the content processingof music. ESPRIT project 28793.


audio descriptors based on the signal (amplitude spectra and parame-ters extracted from the acoustic waveform), music descriptors (musicscores on a symbolic level) and semantic descriptors on the perceptuallevel.

Fig. 4: functional diagram of audio content processing (from CUIDADhttp://www.ircam.fr/produits/techno/multimedia/Cuidad/).

A working model of a description scheme for sound clips and sound effectshas been developed by the American company Musclefishhttp://www.musclefish.com/frameset.html.

The most comprehensive attempt to standardise – among others – the de-scription of audio content is the MPEG-7 effort of the ISO/IECJTC1/SC29/WG11 working group. MPEG-7 is still an ongoing process andthe first version of the standard is expected late in 2001. The following tablegives an overview of the most significant milestones met and the timeline tothe completion of the standard:

• October 16, 1998: Issued formal Call For Proposals• February 1, 1999: Deadline for submission of MPEG-7 proposals.• February 15-19, 1999: MPEG-7 Evaluation Ad Hoc group meeting:


• March 15-19, 1999: Developed the first Experimental Model (XM1.0) and Core Experiments.

• July 6-10, 1999: the first Audio Core Experiments, in Speech Rec-ognition, Sound Effects, Instrument timbre, and Melody, were initi-ated.

• December, 1999: MPEG-7 Working Draft established.• October, 2000: MPEG-7 Committee Draft• February, 2001: MPEG-7 Final Committee Draft• July, 2001: MPEG-7 Draft International Standard• November, 2001: MPEG-7 International Standard• Additional information about MPEG in general and the upcoming

MPEG-7 standard can be found at http://drogo.cselt.stet.it/mpeg/and http://www.darmstadt.gmd.de/mobile/MPEG7/

Although certain details proposed by MPEG-7 may not yet be applicable forlibraries and archives without appropriate applications ready for usage,general guidelines for audio document classification can already be de-rived. Among them are:

• Speech, Speech Recognition Systems• Singing voice• Timbre or Instrument• Instrument Description• Melody, Melody Description• pitch or note (spectrum description)• tempo or rhythm (temporal description)• Surround sound• Sound Effect Classification

Worth mentioning are classification and content description systems origi-nating from musicology as well as from the music industry, the latter beingprogressively present in the commercial download business on the Internet.

4.3 Visualisation of Music Signals

Visualisation of audio signals by so called Spectrograms is employedwhenever music signals cannot be represented in 'textual' formats as musicscore or transcription. This happens frequently in ethnomusicology or whenacoustic and perceptual differences between individual interpretations ofthe same piece have to be documented. Visualisation of music is per-formed in real time so that visual and audio representation of the music canbe observed synchronously. Spectrograms can be read similar to pianorolls comprising the running time axis on the abszissa and the frequencyscale on the ordinate. The strength (level) of spectral components is codedin an appropriate color scale. Spectrogram icons (sound thumbnails) are


used in order to provide fast access to a large number of sound files andsegments stored in a database.

Fig. 5: structuring of audio by visualisation (narrow band spectrogram) of anextract (11 min) from Bruckner 8th symphony.

4.4 Future Development of Content Driven Approaches: Audio Descrip-tion Schemes8

As far as MPEG-7 has already created a structure for “Obvious Audio Ds”four types of audio Ds have been considered:

1.) media based Ds,

2.) non-perceptual low-level audio characteristics,

3.) perceptual low-level audio characteristics and

4.) high-level audio characteristics.

In order to meet best the requirements for the description of audio of keyapplications the group obviously abandoned the demand for generality andconcentrated on the following areas and audio content sets:

8 This section refers to: Obvious Audio Descriptors / Description Schemes; Source:MPEG-7 Audio Reflector; Nov. 1999.


1.) pure music,

2.) pure speech,

3.) pure sound effects and

4.) arbitrary soundtracks applications.

For each of the four application areas it has been provided:

• a typical application scenario in order to prove its relevance,• the effort required to implement it has to be stated,• and a statement whether there’s a chance to automatically determine

the values under consideration.

Currently MPEG-7 is considering the followin Audio Content Set:

Radio A1 Radio news broadcast

Music A2 "Two Ton Shoe" Rock album

A3 Bruckner's Te Deum, and Mozart's Requiem

A4 Original composition, a capella. Voice only

Audio A5 Short sequences of solo instrument and othersounds

A6 Pop song based on an A-A-C motif

According to this pragmatic point of view a step by step creation of usefultools for automatic segmentation and tagging of sound of a large number ofaudio applications is expected. The currently considered descriptors areamong many further possible as the following:

4.4.1 The MPEG-7 Audio Descriptor Scheme (tentative)

Descriptors for pure music:Archiving musicDescriptors for musical genresDescriptors for a composerDescriptors for an artistDescriptors for an artist group (e.g. band, ensemble, orchestra,choir)Descriptors for single pieces of musicSearching music collectionsStructuring music, descriptors to capture musical structures


Filtering music broadcastsMusic education / teachingMusic editingCompositionManipulating musical contentMusic production

Descriptors for pure speech data:Searching speech collectionsStructuring speech collectionsRhethorics education

Descriptors for sound effects:Searching sound effects collectionsMovie synchronisation

Descriptors for arbitrary soundtracks:Searching Radio program collectionsSearching TV program or Movie collectionsFiltering Radio programsFiltering TV programsFilm production / editingFilm education


5. A Modular Archive Model

Long Term Preservation of digital data involves issues of physical storage,software and data standards as well as migration plans and disaster man-agement. In addition, the digital archive involves technology required forglobal, multimedia, object-oriented databases with emphasis on addingvalue along dimensions such as: real-time, fault tolerance, security, andQuality of Service (QoS). The technologies and standards needed to beapplied for the archiving/preservation/retrieval of digital documents is cur-rently a major concern in the archives community. Life time of digital stor-age media and systems is extremly short in comparison to analogue soundand multimedia carriers libraries and archives got used to for many years.The question is whether computer industry will actually provide small andmedium tailored storage solutions for individual libraries in the near future.As an alternative, backup and archive systems located at large computercenters and data farms frequently provide secure digital storage containersand rental storage which could also be used for small and medium volumesof digital data. Librarians and archivists have to force themselves to over-come the psychological barrier not to see a digital document, not to handleit physically anymore, not even to have it in-house – and it is still available,just because it is stored in a secure digital container!

In any case whether the digital storage system is located and managed bythe institution in-house or by a professional computer center remotely thefollowing key functions have to be provided by the archive system (for aReference Model for an Open Archival Information System see: OAIS9:http://www.ccsds.org/red_books.html):

5.1 Acquisition of documents and informationAn entity, which provides the services and functions to accept new docu-ments and adjunct information from external or from internal acquisitionunits under Administration control and prepare the contents for storage andmanagement within the archive. Acquisition functions include receivingsound documents and adjunct information, performing quality assurance onthe document package, generating an Archival Information Package (AIP)which complies with the archive’s data formatting and documentation stan-dards, extracting Descriptive Information from the AIPs for inclusion in thearchive database, and coordinating updates to Archival Storage and DataManagement. Different collections may have different description schemesfitting into the archive’s data formatting and documentation standards.

9 Reference Model for an Open Archival Information System (OAIS): ConsultativeCommittee for Space Data Systems CCSDS 650.0-R-1 RED BOOK May 1999.


5.2 Archival StorageAn entity, which provides the services and functions for the storage, main-tenance and retrieval of AIPs. Archival Storage functions include receivingAIPs from the acquisition unit and adding them to permanent storage, ma-naging the storage hierarchy, refreshing the media on which archive hol-dings are stored, performing routine and special error checking, providingdisaster recovery capabilities, and providing AIPs to Access to fulfill userrequests.

5.3 Data ManagementAn entity, which provides the services and functions for populating, main-taining, and accessing both Descriptive Information - which identifies anddocuments archive holdings - and administrative data used to manage thearchive. Data Management functions include administering the archive da-tabase functions (maintaining schema and view definitions, and referentialintegrity), performing database updates (loading new descriptive informati-on or archive administrative data), performing queries on the data mana-gement data to generate result and query sets, and producing reports fromthese sets.

Fig. 6: Outline of a functional diagram of a modular archive system.

ADMINISTRATIONAcquisition and Access Agreements, IPR Management,

Archive Standards, System Configuration,Physical Access Control, Archival Information Updates,

User Support, User Monitoring.

ACQUISITIONReceiving Information.

Quality Control,Descriptive Information,

Digital Audio WorkstationsCoordination of Updates

DATA MANAGEMENTDatabase Updates,

Database Administration,Query Management,

& Reports, CataloguingDatabase

ARCHIVAL STORAGEReceive Data, Provide Data,

Migrate Media,Management of Storage,

Disaster RecoveryStorage MediaBackup Media

ACCESSDissemination,

Delivery,Data Transfer Control,

& ReportsUser Interface

Music andMusic

InformationProducer

UserCommunity


5.4 AdministrationAn entity, which manages the overall operation of the archive system. Ad-ministration functions include soliciting and negotiating acquisition andaccess agreements with document providers and IPR owners, auditingacquisition material in order to ensure that they meet archive standards,maintaining configuration management of system hardware and software,evaluating the contents of the archive and periodically requesting archivalinformation updates, providing system engineering functions to monitor andimprove archive operations, developing and maintaining archive standardsand policies, providing user support, monitoring changes in the DesignatedUser Communities, interacting with library and archive Management, andactivating stored requests.

5.5 AccessThis entity supports users in determining the existence, description, locati-on and availability of information stored in the archive and allowing users torequest and receive documents and information products. Access functionsinclude communicating with users in order to receive requests, applyingcontrols to limit access to specially protected data and information, coordi-nating the execution of requests to successful completion, generating re-sponses (Dissemination Information Packages, result sets, reports) and de-livering the responses to users.


6. Audio-Networking

6.1 The Role of Libraries and Archives on the Internet

6.1.1 General

The Internet has developed into a mass communications system. By theend of 1998, the Internet had more than 100 million users located through-out the world, and that number is growing rapidly. More than 100 countriesare linked into exchanges of data, news and opinions and more than 1 mil-lion servers are sending information within the net. Obtaining access to in-formation from the net is open to all users who have a personal computeror other access device, the appropriate software and the ability to gain ac-cess to the system (referred to as “obtaining connectivity”), usually providedby an Internet Service Provider (ISP).

Increased availability of bandwidth, faster modems, improved and scalableaudio coding schemes are supporting the development of library music in-formation services. These include the implementation of technology thatallows the digital conversion (digitisation) and storage of mass amounts ofdata as described in previous sections. Future developments might not in-clude a significant change of networt structure but rather a substantial in-crease of capabilities of access devices to download large quantities ofdata; the development of higher bandwidth distribution systems for realtime access and streaming media. The latest research on building, under-standing and using digital archives and digital libraries indicates the devel-opment of sophisticated routers that transmit information; the advent ofuser-friendly software allowing access to information stored on any con-nected computer (search machines and intelligent assets) etc. in the nearfuture.

Libraries and archives have to decide - depending on their regional policies- whether or not they will use the capability to implement

• online services for streaming media,

• services for file transfer or

• services for traditional catalogue access only.

What are the benefits for the library user and what will the consequencesbe for the library organisation and its management? Which are the neces-sary tools and protocols on the Internet and what are the provisions to pro-tect the Intellectual Property Rights (IPRs)?


6.1.2 A Glimps on Copyrights

As libraries and archives can play many different roles for various types oftransmissions over the Internet, it is, consequently, important to examinethese roles separately in determining which activities may give rise to whichliability. The definition, whether an Internet transmission is to be classifiedas

• a communication to the public or as• a communication by telecommunication or as• a process involving reproduction of data, which takes place between

two identifiable partners,

might develop as a central legal issue. Several legislative bodies in differentcountries raised a number of issues including: whether there is a communi-cation by telecommunication to the public as soon as musical sound docu-ment or music information is electronically transmitted, made available,uploaded, downloaded or browsed? Is a communication over a network forwhich access is restricted, a communication to the public? IPR organisa-tions certainly will argue that a communication to the public already occursas soon as the end user can access a library document from a computerconnected to a network.

The role of a library or an archive as content provider is given as soon as

• content is assembled and placed as a collection of fileson a server to allow the files to be accessed.

Usually, the library or archive organisation which has overall responsibilityfor the content of the site (the site owner) also operates and maintains theserver on which the site is located. This model is normally followed bylarger and medium sized libraries and archives (for a detailed descriptionon the role of content providers on the Internet, associated legal issues andbusiness arrangements see e.g. : Public Perfomance of Musical Workspublished by the Copyright Board Canada). For small volumes of digitaldata it is recommended to rent a server as well as storage capacity in a se-cure digital storage container provided by large computer centers.

6.2 Library and Archive Services on the Internet

6.2.1 Connectivity

Digital library services use the Internet as a network of local and remotecomputers and computer networks designed to receive and forward bytesof data grouped into packets between end nodes (the source and destina-tion computers). The basic communication service of the Internet consistsof two components:

• the Internet addressing structure and• the Internet delivery model.


The addressing of computers in the Internet is managed by unique InternetProtocol adresses (IP address). The format of an IP addess is given by acombination of integer numbers, usually in the range of 000 and 255, asfollowing:

The unique IP address allows the control of the World Wide Web traffic.Host names can be allocated according to their geographical location andaccess patterns providing geographical as well as temporal data are ob-tained from each connection. In order to make the addressing easier,slightly more user-friendlier domain names are generally used instead of IPaddresses. These names are translated automatically back to their associ-ated IP addresses by means of the Domain Name System (DNS), operatedby all IAPs for use by their subscribers. Changes of domain names and IPaddresses have to be carried out in co-ordination with the IAP. The domainnames together constitute the Internet’s addressing structure. Once theconnection is established, the service can be initiated providing the appro-priate software is running on the participating computers.

Fig. 7: installing an IP address on a PC under MS Windows. This is not avalid IP address!

6.2.2 Delivery Model

The delivery model contains several delivery modes for the transmission ofmusic and music information with the aim of sending and requesting infor-mation over the Internet. Originally, the network providers tried their best todeliver data but would not provide commitments as to the quality of theservice (QoS, e.g., commitments as to bandwidth or reliability). Generally,the user requests the document required in a unicast pull mode during aconnection. Although it is now possible to request a minimum of requiredbandwidth and a maximum delay (e.g., that packets will be transferredwithin a specified period of time, see Harmonica deliverable 3.6.1 & 3.6.2,RSVP Reservation Protocol), participants in the professional music busi-ness still consider the Internet as inadequate in performance. Moving fromstreaming analogue-based audio transmission to full audio bandwidth pak-ketized digital delivery systems, it is evident that in spite of significant pro-gress in network technology there are many difficulties yet to overcome.Libraries and archives have to decide which of the possiblities they acceptas apllicable for their service.

Alternative delivery modes of providing music and music information overthe Internet involve streaming media as well as multicasting. Real Audio,


currently the market leader for delivering real-time audio and video (includ-ing music)10, provides servers capable to broadcast multiple streams; ne-vertheless, the delivery system for multicast is almost the same as for uni-cast because each recipient still receives its individual copy. Further keyplayers in the real time and music download business are – among manyothers - Liquid Audio, WinAmp and the MP3 community. The selection of aspecific delivery for a music information service has to be decided along thenature of the delivery mode intended to be implemented, whether a directrelationship between the library or archive with the end user can be estab-lished or not. In particular, whether the end user or the library organisationis paying for the usage of the information or the work provided.

6.3 MP3 – An Evolving Digital Music Delivery Sector

WinAmp, Sonique, MusicMatch, RealJukebox, MP3.com, Listen.com, Li-quid Audio, Lycos MP3, 2look4, FileQuest.com, Audiofind, MP3friend,Emusic are just a small selection of sites and search machines of tools andpossibilities for playing, encoding, searching, browsing and downloading ahuge amount of music and music information over the Internet.

A couple of years ago, MP3 was just an audio compression format whichoriginally has been developed by Fraunhofer IIS-A mainly for broacastersprogramme transfer.Today, MPEG Layer-3 is one of the most advancedaudio coding schemes like MPEG-2 AAC (Advanced Audio Coding). In themeantime, MP3 is a growing technology standard for storing and distribut-ing audio on a much broader basis, and is revolutionizing the way audio istransmitted over the Internet. Several manufacturers of audio processingtools and signal processing workstations support MP3 by implementinguser-friendly export options.They feature high quality audio transmissionand support of constant bit rate (CBR) as well as variable bit rate (VBR),encoding at bit rates of up to 320 kbps.

MP3 has become a Net phenomenon that is currently in the center of anenormous controversy. That is because MP3 allows people with an Internetconnection to bypass record stores (and cashiers) and download CD-quality music by their favorite artists - for free. MP3 is not welcomed by mu-sicians and record companies, who expect their sales figures to drop. How-ever, record companies and music publishing houses themselves areadopting this format for promotion purposes and music-on-demand over theInternet. Music industry is still discussing which of the partners in the gamewill be the loosers, which the winners und who will dominate the market?

The home production of MP3-files is easily performed by means of CD rip-pers. CD rippers are programmes that extract - or rip - music tracks from aCD and save them onto the hard drive (Audiograbber). Once the tracks arelocated on the hard drive, they are converted to the MP3 format. Many CDrippers have MP3 encoders built-in (such as MusicMatch Jukebox); or a

10 For a demonstrator system of a digital sound archive seehttp://www.kfs.oeaw.ac.at/DLI/home.html or http://www.sb.aau.dk/Jukebox/edit-report-1.html


separate encoder utility, such as MP3Enc. is needed. Rippers are used tostore the music programme of the current choice on handy hard drives.

For further interesting music sites see: DMX - Digital Music Express ; TCIcable service: 95 music programs;

6.4 Frequently Used Bit Rates

CA*net3 CANARIE–Bell Canada 40 GBit/s optical Internet Oct.1998Internet2 US new national backbone 9.6 GBit/s network for end of 1999Internet2 US new national backbone 2.4 GBit/s network for end of 1998Internet2 US national backbone 22 Mbit/s infractructure April 1998ATM OC-12 622 Mbit/sATM OC-3 155 Mbit/s100-BaseT / FDDI LAN 100 Mbit/sT3 45 Mbit/s10-BaseT Ethernet LAN 10 Mbit/sT1 1.5 Mbit/sDigital HDTV 40-60 Mbit/s 5.1 audio uncompressedNext generation DVD (Blue Laser) 23 Mbit/sDVD-ROM 11.08 Mbit/sDigital DVD-Audio (uncompr.) 9.6 Mbit/s 6 channels max. 96 kHz, 24 bitDigital TV, DVD-Video (NTSC) 6-10 Mbit/s 5.1 audio, NTSC video compr.Multichannel audio compressed 224-640 kbit/s 5.1 channels, Dolby DigitalCompact disc 1.14 Mbit/s 44.1kHz, 16 bit, stereoStereo audio uncompressed 1.536 Mbit/s 48 kHz, 16 bit, stereoStereo audio compressed ("MP3") 20-128 kbit/s MPEG-2 Layer 3Normal telephone channel 64 kbit/s mono, limited bandwidthTelephone modem 14.4- 56 kbit/s ITU V.90 modem 56 kbit/sCable modem (with Ethernet card) 50-200 kbit/s up to 10 Mbit/s theoreticalISDN 64-128 kbit/s FM stereo qualityISDB (Integr.Services Digital Broadc.) 150 Mbit/s NHK trans. 21 GHz/chADSL (a new telephone service) 512 kbit/s uses standard wiresADSL high-speed modem 1 Mbit/sProgram data 10 kbit/sFacsimile (fax) 20 kbit/sStill picture 70 kbit/sTell Text 100-200 kbit/sAudio graphics 800 kbit/s/chMIDI 32.5 kbit/s per 16 channels

Table 2: bit rates currently in use for different audio and multimedia serv-ices (from AES WP-1001 Technology Report TC-NAS 98/1: NetworkingAudio and Music Using Internet2 and Next Generation Internet Capabilitieshttp:/www.aes.org)


7. Appendix

7.1 A Sample List of Audio Players

1.00 for Macintoshhttp://www.hitsquad.com/smm/programs/Little_Player/

a2b Music Player 1.00b8 for Macintoshhttp://www.hitsquad.com/smm/programs/a2b_mac/

a2b Music Player 2.0 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/a2b/

ARIES Breathe MP3 player 0.91 for Windows 95http://www.hitsquad.com/smm/programs/ARIES_Breathe_MP3_player/

Aries Mod Player 1.0 for Windows 95, NT 4.0http://www.hitsquad.com/smm/programs/Aries_Mod_Player/

Audio CD Player for Windows 3.1http://www.hitsquad.com/smm/programs/Audio_CD_Player/

Audioactive Player 1.2a for Macintosh PPChttp://www.hitsquad.com/smm/programs/Audioactive_Player_PPCMac/

Audioactive Player 1.3 for Windows 3.1http://www.hitsquad.com/smm/programs/Audioactive_Player_win3/

Audioactive Player 1.9 Beta for Windows 95/NThttp://www.hitsquad.com/smm/programs/Audioactive_Player_win95/

Beatnik Player 2.03 for Macintoshhttp://www.hitsquad.com/smm/programs/Beatnik_mac/

Beatnik Player 2.03 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Beatnik/

CD player Maximus 3.3 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/CDMaximus_win32/

Cubic Player for Doshttp://www.hitsquad.com/smm/programs/Cubic_Player/

Destiny Media Player 1.31 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Destiny_Media_Player/

Digital Music Player for OS/2http://www.hitsquad.com/smm/programs/Digital_Music_Player/

DSM Player 1.04 for Macintoshhttp://www.hitsquad.com/smm/programs/DSM_Player/

Dual Module Player for OS/2http://www.hitsquad.com/smm/programs/Dual_Module_Player/


Hyperprism H-PPC Player for Macintoshhttp://www.hitsquad.com/smm/programs/Hyperprism_H-PPC_Player/

iNERTiA PLAYER for Doshttp://www.hitsquad.com/smm/programs/iNERTiA_PLAYER/ ,

Liquid Player 5.0 Preview for Macintosh http://www.hitsquad.com/smm/programs/LiquidAudio_mac/

Liquid Player 5.0 Preview for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/LiquidAudio/

Melody Player 2.0 for Windows 95http://www.hitsquad.com/smm/programs/Melody_Player/

Microsoft Windows Media Player 6.4 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/MicrosoftMediaPlayer/

MIDI Player V1.55 for Windows 95http://www.hitsquad.com/smm/programs/MIDI_Player/

Midi Synthi Player 5.8 for Windows 95/98/3.1http://www.hitsquad.com/smm/programs/Midi_Synthi_Player_win95/

Midisoft Internet Media Player v3.08 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Internet_Media_Player/

Mikey Player 98 4.1 beta for Windows 95http://www.hitsquad.com/smm/programs/Mikey_Karaoke_Player/

Mikeys mp3 Player for Windows 95, 98http://www.hitsquad.com/smm/programs/Mikeys_mp3_Player/

Mini MIDI Player v1.12 for Windows 95http://www.hitsquad.com/smm/programs/Mini_MIDI_Player/

MM Player 4.02 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/MM_Player/

MM Player Pro 4.02 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/MM_PlayerPro/

MODPlug Player 1.40 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/ModPlug_Player/

Mpeg Audio Player 1.21 for Macintoshhttp://www.hitsquad.com/smm/programs/Mpeg_Audio_Player/

Multi Module Music Player 1.00b4a for Windows 95http://www.hitsquad.com/smm/programs/Multi_Module_Music_Player/

Musician's CD Player for Windows 95http://www.hitsquad.com/smm/programs/Musicians_CD_Player/

NoteWorthy Player 1.50 for Windows 3.1http://www.hitsquad.com/smm/programs/NoteWorthy_Player_win3/


NoteWorthy Player 1.55a 16 bit for Win 3.xhttp://www.hitsquad.com/smm/programs/NoteWorthy_player_16/

NoteWorthy Player 1.55a for Windows 95/NThttp://www.hitsquad.com/smm/programs/NoteWorthy_Player_win95/

PH Player for Atarihttp://www.hitsquad.com/smm/programs/PH_Player/

Player PRO Direct-To-Disk 0.1b for Macintoshhttp://www.hitsquad.com/smm/programs/Player_PRO_Direct-To-Disk/

PROcessu CD Player 2.02 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/PROcessu_CDPlayer/

Real Player G2 Update 1 for Windows 95http://www.hitsquad.com/smm/programs/RealPlayerG2_win32/

RealPlayer 5.0 for Windows 3.1http://www.hitsquad.com/smm/programs/RealPlayer_win3/

RealPlayer G2 for Windows 95http://www.hitsquad.com/smm/programs/RealPlayer_win32/

Shockwave 6 Flash Player for 68K for Macintoshhttp://www.hitsquad.com/smm/programs/Shockwave_Player_68K/

Simple CD Player 2.3 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Simple_CDPlayer/

SSEYO Koan File Player V2.2 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/SSEYO_Koan_FilePlayer/

Streaming Audio Player 0.8 beta for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/SAP/

ThrottleBox Player 1.2 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/ThrottleBox_Player/

TitleTrack CD Player v2.1 for Macintosh PPChttp://www.hitsquad.com/smm/programs/TitleTrack/

True Speech Audio Player for Mac (PPC) for Macintosh http://www.hitsquad.com/smm/programs/True_Speech_Audio_ppc/

True Speech Audio Player for Mac 68k for Macintoshhttp://www.hitsquad.com/smm/programs/True_Speech_Audio_mac68/

Ugly CD Player 2.1 for Macintoshhttp://www.hitsquad.com/smm/programs/Ugly_CD_Player/

Unreal Player Max 2.02 for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Unreal_Player/

Upscale Pro MIDI Player for Windows 95http://www.hitsquad.com/smm/programs/UpscaleProMIDI/


Variable Speed CD Player for Windows 95http://www.hitsquad.com/smm/programs/variablespeed/

Wired Planet player for Windows 95/98/NThttp://www.hitsquad.com/smm/programs/Wired_Planet_player/

Xing MP3Playerhttp://www.xingtech.com/mp3/player/

ya cd player 2.5 for Macintoshhttp://www.hitsquad.com/smm/programs/ya_cd_player/

Yo!MPEG Player v1.0.2.79 for Windows 95/98http://www.hitsquad.com/smm/programs/YoMPEG_Player/