INTERNATIONAL ORGANIZATION FOR STANDARDIZATION
`ORGANISATION INTERNATIONALE DE NORMALISATION
`ISO/IEC JTC1/SC2/WG8
`CODED REPRESENTATION OF PICTURE AND AUDIO INFORMATION
`
`Source
`
`Title
`
`Status
`
`Hans-Georg Musmann
`Chairman of MPEG/Audio
`Ninth MPEG Meeting Report
`Separate Meeting of the MPEG/Audio Group
`Draft
`
`ISO/fEC JTC1/SC2/WG8
`MPEG 88/237
`December1989
`
`Ninth MPEG Meeting Report
`Separate Meeting of the MPEG/Audio Group
`
`On December
`11 and 12 1989 the MPEG/Audio group met at
`the Swedish Broadcasting
`The list of participants of this meeting is given
`Corporation Sveriges Radio in Stockholm/Sweden
`The meeting agenda is given in Annex II
`in Annex
`
`The Yokosuka Kurihama meeting report was approved
`
`Following the request of several group members to discuss the impacts of
`Systems group agreed on during the Kurihama meeting document MPEG 89/216 also included
`in the Proposal Package Description MPEG 89/128
`paragraph 6.1 the proposed agenda was
`extended by the item System aspects which was allocated between agenda items
`and
`
`the conclusions
`
`the
`
`The chairman informed the group about activities
`since the October meeting in Kurihama/Japan
`The EBU indicated its interest
`to co-operate with the MPEG/Audio group and offered technical
`support for the subjective testing document MPEG 89/227 Thames Television confirmed its offer
`to perform the subjective tests in case the Swedish Broadcasting Corporation is unable to proceed
`document MPEG 89/229
`
`By November17 the following three revised versions of the detailed descriptions of sound coding
`algorithm were submitted
`
`BTRIJNTT document MPEG 89/207 Rev
`
`Fujitsu/NEC/JVC/Sony
`
`document MPEG 89/208 Rev
`
`Matsushita
`
`document MPEG 89/210 Rev
`
`in the document collection of the Japan meeting
`These revised versions were already included
`which meanwhile was distributed to all MPEG members
`
`1989 the two remaining individual algorithm proposers CNET and Matsushita
`By December
`that they have joined the cluster corresponding to their algorithm strategy so that there
`declared
`are only the following four development groups left
`
`1
`
`HP 1030
`
`

`

`Transform Coding with overlapping blocks
`ATT CNET DTB RIG
`
`Transform Coding with non-overlapping blocks
`Fujitsu JVC NEC Sony
`
`Subband Coding with less than or equal to
`BT NTF
`
`subbands
`
`Subband Coding with more than
`CCETF IRT Matsushita Philips
`
`subbands
`
`no preselection is necessary The detailed descriptions of the individual algorithms
`Therefore
`documents MPEG 89/209 and 89/210
`of CNET
`have been withdrawn Due to the contribution
`to the proposal of development group
`in the above list there are some amendments to the
`detailed description of the algorithm called ASPEC document MPEG 891205 which are
`contained in document MPEG 89j234
`revised version of document MPEG 89/205
`can be
`in document MPEG 89/234
`assembled by replacing and adding the respective
`pages as explained
`Development group
`in the above list apologized for not being able to submit an update of the
`detailed description including the contributions of Matsushita because the negotiations between
`could be finalized only just before this meeting The intention of submitting this
`the companies
`in March 1990 as indicated in letter by Mr Kogure
`update at the next meeting at Tampa/Florida
`document MPEG 89/235
`could not be accepted by the group Therefore
`it was agreed that this
`update has to be provided at latest onDecember3 11989 In the meantime this update was received
`and sent out by the chairman Furthermore it was agreed that
`there should not be any temporal
`limitation for addressing
`to the originator concerning the algorithm description
`questions
`
`As requested by the audio group at
`the Yokosuka meeting the chairman of the VLSI ad hoc
`group provided some preliminary
`information about multiplier and adder speed of operation
`document MPEG 89/226
`similar procedure as was used for getting the VLSI questionnaire
`results documents MPEG 89/120 and MPEG 89/172 will be applied to get final
`results for these
`operation speeds which then will be included in revised version of document MPEG 89/172 This
`revised version should be available at latest at the Tampa meeting in March 1990 Mr Chiariglione
`the end of January 1990 Concerning the
`try to provide further preliminary
`information at
`silicon area the table of document MPEG 89/172 will be applied in the present state
`
`will
`
`If the hardware to be used for the final
`tests is based on floating point arithmetic
`then the
`computation of encoder and decoder costs ancLor complexity has also to be based on floating point
`arithmetic If the hardware is based on integer arithmetic
`then the computation of encoder and
`decoder costs and/or complexity has also to be based on integer arithmetic
`
`The decoder
`is expected to be an universal decoder
`bitrate The encoder may be different for every bitrate
`
`that is the decoder
`
`is the same for every
`
`For testing purposes the encoder output has to be
`and
`continuous clock
`
`bit by bit interleaved serial data bitstream
`
`The error insertion can be done either by
`separate error insertion box outside the decoder with
`the problem that there must be one such box for every system to be tested or in way proposed by
`Mr Dehery CCEIT where every proposer has to provide very simple hardware at the decoder
`input as illustrated in diagram in Annex musing random noise generator provided by the BBC
`
`2
`
`

`

`decision between these two alternatives
`Broadcasting Corporation has to be made at
`
`depending mainly on the requirements of Swedish
`the next meeting
`
`of
`
`The discussion about System aspects was based on some items of the Kurihama meeting report
`the Systems ad hoc group document MPEG 89/216
`also included
`in the Proposal Package
`Description MPEG 89/128 paragraph 6.1 Concerning the aspect of fast forward audio playback
`items 21 and 22 of the Systems report it was agreed that
`the definition in paragraph 73.2 of the
`proposal package description document MPEG 89/128 is for testing purposes The problem of
`item 25 of the Systems report will
`as agreed on former meetings not be considered
`transcoding
`by the audio group
`
`11 and 13 of the Systems report were
`The aspects of random access and audio editing items
`discussed again very thoroughly It was realized that the length of an access unit is not the relevant
`for scoring the performance of random access access unit is defined
`to be measured
`parameter
`as the smallest part of the audio signal which can be decoded by itself where decoded means fully
`sound.Therefore anew parameter called addressable unit was defined as follows
`reconstructed
`
`An addressable unit is the maximum distance between entry/exit points in the encoded
`bitstream expressed in ms of the decoded sound signal
`
`An addressable unit will contain
`fixed number of bits which represent
`that for the input sampling rate of 44.1 kHz
`samples with the exception
`difference of
`allowedin paragraph 73.3 of the proposal package description document MPEG 89/128
`every
`ofaccess unit was substituted by addressable unit and the parameters for evaluation
`appearance
`of the corresponding score were modified Furthermore the requirement that an addressable unit
`has to be less than 1/30
`was put on the requirement
`list in paragraph 6.3 of the proposal package
`description document MPEG 89/128 The length of an access unit will now be only considered
`list and has to be less than 100 ms All paragraphs of the proposal package
`in the requirement
`description document MPEG 89/128 where changes were made can be found updated in Annex
`Iv
`
`fixed number of
`
`bit is
`
`After having studied the related MPEG/Audio documents
`the experts of Swedish Broadcasting
`they will be able to perform the sound quality assessment
`Corporation decided that
`tests in
`Stockholm Mr Ryden Mr Grewin and Mr Bergman gave
`brief summary of their discussions
`in document MPEG 891232 and introduced the available equipment which is
`as described
`in documents MPEG 89/231
`and 891232 According to the request of Swedish Broad
`described
`casting Corporation that the subjective assessment should be based upon hardware as explained
`in
`of document MPEG 89j232 it was agreed that hardware encoder
`decoder will be
`paragraph
`provided for the subjective tests Concerning the problem of D/Aconverters
`it is expected that two
`enough STAX Lambda Pro
`can be found The problem of providing
`identical DIA converters
`headphones will be solved within the EBU
`
`task force for elaborating and finalizing the details of the test procedures
`It was decided to setup
`and test evaluation This task force will consist of
`the Swedish Broadcasting
`of
`the experts
`Corporation optionally one or two experts of the EBU and mandatory one representative of each
`this task force is Mr Thomas Ryden of
`development Group The chairman of
`the Swedish
`Broadcasting Corporation Each of
`groups is asked to nominate
`the four development
`one
`the second week of January 1990 that
`latest on Friday January 12 1990
`representative until
`is at
`this task force will be settled the further actions will be conducted
`to support this task force After
`by Mr Ryden whose address is as follows
`
`3
`
`

`

`Thomas Ryden Sveriges Radio Technical Department S-lOS 10 Stockholm Sweden
`Tel 46-8-784-6509Fax
`46-8-662-6546
`
`In order to reduce the number of
`procedure was agreed
`
`test sequences
`
`to be used in the final
`
`tests the following
`
`The 20 agreed test sequences document MPEG 89/166 will be kept until the final
`
`tests
`
`Test
`
`tapes with processed data of all 20 test sequences
`at the bitrates of 64 96 and 128
`to the state of the algorithm at
`Kbit/s corresponding
`the end of May 1990 have to be
`distributed until June
`1990 as follows each development group has to provide every
`other development group and the Swedish Broadcasting Corporation with one test
`tape
`
`directly
`
`After having analyzed these test
`tapes every development group may give recommenda
`tions for the selection of the most critical
`test sequences
`including indications of weakness
`points of system These recommendations have to be submitted to the Swedish Broad
`casting Corporation until June
`1990 and will not be distributed to any other person It
`is up to the Swedish Broadcasting Corporation wether and how these recommendations
`will be considered or ignored in the selection of
`reduced set of test sequences
`for the
`final tests
`
`The Swedish Broadcasting Corporation is free to select additional unknown critical
`and to fix the number of test sequences
`sequences
`to be used in the final
`
`tests
`
`test
`
`Each of the four development groups is allowed to nominate up to
`subjects to participate in
`the subjective tests Residual subjects will be selected by the Swedish Broadcasting Corporation
`so that the total number of subjects will be approximately 50
`
`In letter submitted by Mr Taka NiT document MPEG 89/228
`it was proposed to reconsider
`the time schedule to allow period of collaborative improvement of the selected algorithm This
`proposal was discussed thoroughly and it was realized that perhaps
`better solution would be
`possible by adopting this procedure but then it would not be possible to keep the time schedule
`resulting in draft proposal in September
`1990 It was agreed that it would be better to come up
`with
`good solution at
`the defined
`than to achieve
`deadline
`slightly better solution sometime in
`the future Therefore
`it was decided to stick to the former agreed time schedule Nevertheless
`more detailed specification of the time schedule was discussed see also Annex IV
`
`The next meeting will be during the WG8 meeting in Tampa Florida USA from March 15 to
`March23 1990 The MPEG/Audio group wifi meet from March 19 to March21 but probably there
`will be special joint meetings together with the System- and VLSI ad hoc groups during the other
`days Especially on March 19 there will be
`joint meeting of the Audio Video and Systems group
`
`It was agreed to distribute updated descriptions of algorithms which are reflecting the state of
`the algorithm at
`the end of February 1990 together with test
`tapes containing
`processed test
`corresponding to the submitted updated descriptions before the next meeting in order
`sequences
`to give all development groups and the Swedish Broadcasting Corporation more information about
`the quality to be expected and the changes in the algorithm Each development group has to provide
`every other development group and the Swedish Broadcasting Corporation with one test
`tape
`directly The test
`tapes have to be sent out by express mail DI-IL at latest on March
`1990 The
`content of the test
`tapes has to be as follows
`
`4
`
`

`

`all 20 test sequences
`
`processed at
`
`bitrate of 96 kbit/s mandatory
`
`some or all 20 test sequences processed at
`
`bitrate of 128 kbit/s optionally
`
`some or all 20 test sequences processed at
`
`bitrate of 64 kbit/s optionally
`
`those
`
`in paragraph 7.3.8 of the proposal package descripion
`selected
`test sequences
`bitrate of 96 kbit/s with inserting random errors bit error rate of 10-
`in
`processed at
`the encoded bitstream The margin for ancillary data 1.5 kbit/s has not to be used for
`coding bits or synchronization mandatory
`
`As described
`
`tapes have to be distributed until June
`in detail on the previous page the final
`test
`have to be submitted until
`1990 and recommendations
`for the final selection of test sequences
`
`June
`
`1990
`
`The final detailed descriptions of algorithms have to be submitted to the chairman at
`latest on
`June 18 1990 The encoder and decoder hardware has to be ready and available for the Swedish
`Broadcasting Corporation on June 18 1990
`
`of document MPEG 89/232 the Swedish Broadcasting Corporation proposes
`In paragraph
`schedule for the testing period which is separated in three phases
`
`Preparation phase
`
`subjective assessments and
`
`evaluation
`
`phase
`
`one engineer has to be present during the whole time from the
`For every system to be tested
`the end of the subjective assessments The preparation
`beginning of the preparation
`phase until
`phase will start on June 18 1990 and the subjective assessments will start on July
`1990
`
`In order to avoid that the subjects have to spend too much time in Stockholm it was proposed
`into two parts and to involve only half of the test subjects
`to split the phase of subjective assessments
`in each part In the first part the systems could be tested at the bitrates of 128 and 96 Kbit/s only
`and in the second part at 12896 and 64 Kbitls as well as the performance of fast forward and error
`
`sensitivity
`
`Finally there were two requests related to the agenda item A.O.B CNET submitted
`document
`MPEG 891233 concerning the importance of 64 kbit/s
`for ISDN applications and proposed an
`procedure to verify that the chosen algorithm is suitable to ISDN require
`additional
`evaluation
`ments The necessity of applying such
`procedure was not realized
`
`Mr Petit CNET raised another question concerning the possibility of providing other stand
`tests The general opinion was that
`ardization bodies with the results of
`this should be
`possible But then the full results together with the rules of assessment should be distributed
`at this meeting the answer was postponed until
`one development group was not present
`the next
`it was recommended that Mr Petit submits his request in written form as
`meeting Additionally
`an official document
`
`the final
`
`Since
`
`5
`
`

`

`Annex
`
`List of Participants MPEG/Audio
`
`Stockholm 11-12 December 1989
`
`Schroeder
`
`DTB Deutsche Thomson-Brancit GmbH
`
`Gerhaeuser
`
`FhG Fraunhofer Gesellschaft
`
`Meier-Engelen
`
`DIR German Aerospace Research Establishment
`
`NoH
`
`Musmann
`
`Fuchs
`
`Stoll
`
`Dehery
`
`Mahieux
`
`Petit
`
`Komly
`
`Gilchrist
`
`Kim
`
`Chiariglione
`
`Carter
`
`van de Kerkhof
`
`de Wit
`
`Bergman
`
`Grewin
`
`Rydn
`
`Brandenburg
`
`Universitaet Berlin
`
`Universitaet Hannover
`
`Universitaet Hannover
`
`IRT
`
`Institut
`
`flier Rundfunktechnik GmbH
`
`CCETF
`
`CNET France Telecom
`
`CNET France Telecom
`
`1DF Television de France
`
`BBC British Broadcasting Corporation
`
`BT
`
`British Telecommunications
`
`plc
`
`CSELT
`GO Graphics Communication Technologies
`
`Philips CE
`
`Philips CE
`
`Swedish Broadcasting Corporation
`
`Swedish Broadcasting Corporation
`
`Swedish Broadcasting Corporation
`ATT Bell Laboratories
`
`GB
`
`GB
`
`NL
`
`NL
`
`USA
`
`6
`
`

`

`Annex II
`
`Meeting Agenda
`
`Stockholm 11-12 December 1989
`
`MPEG/Audio Group
`
`Opening of the meeting
`
`Approval of the agenda
`
`Allocations of contributions
`
`Communications
`
`from the chairman
`
`Yokosuka meeting report
`
`Decision about preselection
`
`6A
`
`System aspects
`
`Finalizing of subjective test procedures
`
`Forward audio playback A-B-X test procedure paragraph
`
`72.1 of PPD
`
`Fast forward audio playback anchor paragraph
`
`7.2.2 of PPD
`
`Intermediate
`
`audio quality lower anchor paragraph 7.2.7 of PPD
`
`Robustness to bit errors lower anchor paragraph 7.2.8 of PPD
`
`Preparation of subjective tests
`
`Preparation
`
`of test
`
`tapes
`
`Selection of test subjects
`
`Test equipment
`
`Revision of time schedule
`
`Recommendations
`
`of the Stockholm meeting
`
`AO.B
`
`Close of the meeting
`
`10
`
`11
`
`12
`
`7
`
`

`

`Annex ifi
`
`Proposal for Error Insertion Interface Mr Dehery
`
`Clock
`
`DCK
`
`DEC
`
`PRIN
`
`Check
`
`PRIN Clock
`it
`
`Check\
`
`___//
`
`______I
`DB
`
`Clock
`
`Random Generator
`
`Clock
`
`PRIN
`
`8
`
`