NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/ai/4274

Internet Message Format | 1992-11-14 | 17.5 KB

Xref: sparky comp.ai:4274 comp.sys.hp:12890 comp.sys.mac.apps:17173
Newsgroups: comp.ai,comp.sys.hp,comp.sys.mac.apps
Path: sparky!uunet!pageworks.com!world!eff!sol.ctr.columbia.edu!usc!elroy.jpl.nasa.gov!ucla-cs!oahu.cs.ucla.edu!shank
From: [e-mail] (Eric Shank)
Subject: Re: Info sought on Optical Character Readers (for music)
Message-ID: <[e-mail]>
Sender: [e-mail] (Mr Usenet)
Nntp-Posting-Host: oahu.cs.ucla.edu
Organization: UCLA, Computer Science Department
References: <[e-mail]>
Date: Fri, 13 Nov 92 19:59:27 GMT
Lines: 368

Regarding Optical Music Recognition-

This is somewhat of a frequently asked question; following is a
summary of articles I've collected from the net over the years, to
which I would add that two theses (one old PhD and one recent MS have
been done at MIT on the subject; check your local library)


Following articles are both questions and answers, reposted without
authors permission:
__________

Date: 1 May 90 22:45:27 GMT
From: Torkil Hammer <[e-mail]>
Subject: optical music recognition
To: music-research@prg
Message-ID: <[e-mail]>

I have a large collection of printed sheet music and want to convert
it into machine readable form through a scanner.  Is there any software
available for doing the "optical character recognition" of the notes?

Torkil Hammer

------------------------------
Date: 19 Oct 90 16:31:42 GMT
From: Dave Baines <[e-mail]>
Subject: Optical musical score recognition
To: music-research@prg
Message-ID: <[e-mail]>

I am posting this for a student here who does not have access to UseNet so
please send any replies directly to him via E-mail.

------------------------------------------------------------------------------

Optical Musical Score Recognition

  I am seeking information for my final year honours project at the department
of Computer Science, Edinburgh University.  The project involves interpreting
musical features (notes, sharp signs etc) from an image produced by an optical
scanner.  Has anyone ever tackled a similar project, or know of any literature
in this field ?

Thanks in advance,
David Bainbridge ([e-mail])

------------------------------

Date: 21 Oct 90 04:43:28 GMT
From: John M Davison <davisonj%en.ecn.p[e-mail]>
Subject: Optical musical score recognition
To: music-research@prg
Message-ID: <[e-mail]>

    The WABOT-1 musical robot, which is pictured on the cover of
the Spring 1986 _Computer_Music_Journal_ and written up in the Summer
1986 _Computer_Music_Journal_, could (unless I am mistaken) scan a
page of sheet music in about ten seconds and convert the score to a
sequence which WABOT-1 would subsequently play.  (The sheet music was
scanned in one shot; real-time scanning, such as a human performer
would do, was not implemented.)

[e-mail]

------------------------------

Date: Sun, 21 Oct 90 09:03:11 BST
From: Stephen.Page@prg
Subject: Optical musical score recognition
To: [e-mail]
Cc: music-research
Message-ID: <[e-mail]>

I'm not aware of any recent work in this field (it seems to have died away...)
but here are the basic references for foundation work in the field:

Clarke, A.T., B.M. Browne, and M.P. Thorne. "Inexpensive Optical Character
Recognition of Musical Notation: A New Alternative for Publishers." In
_Computers_and_Music_Research_. Proceedings of a conference held on 11-14 April
1988. [Ed. Alan Marsden.] Lancaster: Centre for Research into the Applications
of Computers to Music, Univ. of Lancaster, [1988], pp. 84-87.

Kassler, Michael. "An Essay Towards Specification of a Music-Reading Machine."
In _Musicology_and_the_Computer_. Ed. Barry S. Brook. New York: City Univ.
of New York Press, 1970, pp. 151-175.

Kassler, Michael. "Optical Character-Recognition of Printed Music: A Review
of Two Dissertations." _Prespectives_of_New_Music_, 11, No. 1 (Fall-Winter
1972), p. 250.

Prerau, David S. "Computer Pattern Recognition of Printed Music." _AFIPS_
Conference_Proceedings_, 39 (1971), 153-62.

Prerau, David S. "Computer Pattern Recognition of Standard Engraved Musical
Notation," Diss. MIT. 1970.

Prerau, David S. "DO-RE-MI: A Program that Recognises Music Notation.
_Computers_and_the_Humanities_, 9 (1975), 25-29.

Roads, Curtis. "The Tsukuba Musical Robot." _Computer_Music_Journal_, 10,
Mo. 2 (Summer 1986), 39-43.


Here's an excerpt from my dissertation (Stephen Dowland Page, "Computer Tools
for Music Information Retreival," Diss. Oxford 1988, pp. 43-44):

"The use of a device which can scan a printed score is, in theory, the most
effective method of music input. It allows exact, uninterpreted input of the
written score. [Previously I had discussed the differences between written,
performed, and perceived music.]

Michael Kassler set out an abstract specification for a "music-reading machine"
in 1970. He defined the input (a certain vocabulary of musical symbols) and
output (a stream of binary digits) of a machine he called "M". Although this
is an interesting exercise, "M" has never been built and probably will not be
built for many years. There are many practical problems to be solved at a
lower level--for example, the recognition of simple note shapes.

Although optical scanning of printed letterpress is at an advanced stage,[1]
music has many more complexities than text, such as the variable shape of
slurs, the wide variety of music fonts, and the two-dimensional nature of a
score, which must be recognised by a machine. David Prerau, in his 1970
dissertation, put forward optical methods which have been successfully
tested on a small fragment of two-part music. A recent Japanese experiment,
a "musical robot" which can play music on a keyboard, incorporates optical
scanning. Beyond these few projects, development of optical scanning has
been very limited.

[1] The Kurzweil Data Entry Machine has been in use in literary studies for
a number of years. After "training" on a font it can read pages of text into
the computer faster, and more reliably, than a person can type them. Already
this machine has been made obsolescent by faster, more 'intelligent'
machines."

This is a tricky area. Watch your scope!
By the way, Michael Kassler is now connected to an email network, as
    [e-mail] .


                                                - Dr Stephen Page
------------------------------------------------------------------------------
4 Harwood Terrace             Technology and Systems Integration Division
Fulham                        Andersen Consulting
London    SW6 2AB             2 Arundel Street
U.K.                          London    WC2R 3LT
                              U.K.

JANET   [e-mail]                    | Work 071-438 5000
ARPA    [e-mail] |   [phone]
UUCP    ...!ukc!ox-prg!sdpage                      | Home 071-371 9981
BITNET  [e-mail]              |   [phone]
        Fax: [phone]    Telex: 8812711

------------------------------
Music-Research Digest       Sat, 27 Oct 90       Volume 5 : Issue  88

Today's Topics:
                         ANSI X3V1.8M update
              Optical musical score recognition (2 msgs)


*** Send contributions to [e-mail]
*** Send administrative requests to Music-Research-Request

*** Overseas users should reverse UK addresses and give gateway if necessary
***     e.g.   [e-mail]
***     or     [e-mail]

*** Back issues, index, etc.: send "help" in a message to archive-server
***   @uk.ac.oxford.prg (in the UK) or @hplpm.hpl.hp.com (elsewhere)

----------------------------------------------------------------------

Date: Wed, 24 Oct 90 12:04:37 EDT
From: "Steven R. Newcomb" <[e-mail]>
Subject: ANSI X3V1.8M update
To: Music-Research <[e-mail]-relay>
Message-ID: <[e-mail]>

This note is an update on the ANSI X3V1.8M Standard Music Description Language
(SMDL) activity, and it is prompted by Edward L. Stauff's query in MRD 5.86.

>   1.  How (if at all) can I get a copy of the spec?  (ANSI in NYC doesn't
>       seem to know about it.)

The committee no longer distributes the working drafts of the standards
in electronic form for various reasons.  The drafts available from the
MRD archives are hopelessly obsolete and they should probably be
discarded (or heavily disclaimered) to minimize confusion.  Printed
copies of the most recent draft is always available from the X3V1.8M
Secretariat, The Computer Music Association, Larry Austin, President, P.
O.  Box 1634, San Francisco, California 64101-1634 USA.  I don't know
what they are charging these days per page, but there will be a
duplicating, handling, and postage charge.  If all you are interested in
is technical information, then what you want is two documents:
X3V1.8M/SD-7 (ANSI Project X3.749-D, "HyTime", which is the music and
time-based document time and hyperlink standard, about 80 pp.), and
X3V1.8M/SD-8 (ANSI Project X3.542-D, "Standard Music Description
Language [SMDL]", which is everything about music except for what is in
SD-7, about 60 pp).  In addition, I would urge you to consider ordering
X3V1.8M/SD-0, our General Information and Guide to Participation
document (4 pp); X3V1.8M/SD-2, our Document Register (a list of
committee documents, most of which are also available from the
Secretariat); X3V1.8M/SD-3 (list of current committee participants);
X3V1.8M/SD-6 (User Needs and Functional Specification).

Please be informed that all of these documents are works in progress and
they can all be expected to change.  The next meeting will be in San
Jose, California, Nov 12-16, 1990.  The meeting announcement and draft
agenda can be found in document X3V1.8M/90-66.

It is not particularly astonishing that whoever Mr.  Stauff talked to at
ANSI does not know about X3V1.8M, since it functions at a leaf node of
the ANSI organization, which is vast and variegated.  Perhaps if he had
mentioned the project numbers, they would have been more helpful.  Thank
you for bringing this glitch to my attention; I'll talk to them and
bring them up to speed on this.

>   2.  Has it become stable enough that it's worth starting to develop tools
>       to support it?

I would not recommend large investments in SMDL tools per se as yet,
since the standard is not complete.  It's been fairly stable for a year,
but that does not mean that some major unforeseen problem won't force
some parts to be redesigned between now and its adoption as an ANSI
and/or ISO standard.  TO ALL WHO ARE INTERESTED IN BUILDING SMDL TOOLS:
please obtain the documents listed above, read them carefully, and give
us your constructive criticism and suggestions; contributing written
responses to the documents makes you an official participant in the
X3V1.8M standard-building process.  Your input is vital to the technical
success of SMDL.

>   3.  Is anyone currently developing tools for it (i.e. editor, print
>       formatter, conversions, etc.)?

At the Florida State University Center for Music Research we intend to
build tools, but we're not doing it yet, because we are concentrating
our efforts on refining the standard itself.

Although I cannot speak on their behalf, many people and organizations
have expressed their eagerness to apply HyTime and SMDL to various
problems.  The array of applications turns out to be staggeringly
large.  Even the U. S. Air Force has expressed great interest, believe it
or not.

One application particularly exciting to me is representing scores which
have been read by an automatic music-reading system.  Dr.  Nicholas
Carter of the University of Surrey has expressed interest in adapting
his evolving system for automatic score reading in such a way as to
output SMDL.  (Dr.  Carter is working with the Oxford University Press
on this project.  Incidentally, I was surprised that his project, which
is the most advanced one I know of, was not mentioned in the recent
rundown of score reading projects here in MRD.)

>   4.  Is there any indication that the MIDI software industry is going to
>       support it?  That is, can I expect programs like SCORE, FINALE,
>       COPYIST, etc. to support it someday?

Dave Oppenheim of OpCode Systems is a longtime participant; he has kept
us abreast of MIDI files standards so that we can avoid leaving out any
of the information in that information set.  We have not heard from Dr.
T's, and we do not know what their plans are.  Dave Kusek, President of
Passport Designs, which markets the Score product, was actually the host
of the first meeting of X3V1.8M (1986, Half Moon Bay, California).
Passport has not chosen to send any representatives to meetings
recently, but they are well aware of our activities.  Coda Music
Software, which markets the Finale product, was an enthusiastic
supporter of X3V1.8M, and it hosted a meeting in Minneapolis (1987).
Coda remains interested and aware; I know this because they were
accidentally deleted from the mailing list (when their usual
representative to the committee left the company) and we were asked to
place their CEO's name on our list as his replacement.  To summarize, I
would say that the prospects for universal (or near-universal) support
of SMDL are excellent, but for now we must focus on making the standard
technically excellent.  The rest will follow naturally.  Other music
industry participants: American Interactive Media (Philips-Sony); Human
Touch; Yamaha International; MIDIWORLD; Warner New Media; Interactive
Media Systems; Graphire Corporation.

The HyTime ("everything but pitch and meter") aspect of the project has
attracted considerable interest; here are some of those industrial
participants and observers: National Institute of Standards and
Technology (NIST); Mead Data Central; Optical Publishing Association;
General Electric; SGML Associates; Hewlett-Packard; Search Technology;
IBM; Bellcore; Xerox; Army Research Institute; Central Intelligence
Agency; CCETT; Bruel & Kjaer; Digital Equipment Corp; Oak Ridge National
Laboratory; APL Integrators; ANSER Analytic Services; Knowledge
Management Associates; Apple Computer; Oster & Associates; Industrial
Technology Institute; Information Navigation; SoftQuad, Inc.; Mitre
Corporation; Tektronix Labs; Bull HN Information Systems; Sun
Microsystems; Taunton Engineering; Texas Instruments; Planning Analysis
Corporation; Context Corporation.

> Any information (including opinions, speculation, etc.) would be appreciated.
> My interest in SMDL is in the possibility of using it as a vehicle for
> exchanging music (scores) across the network.

Because SMDL is an application of Standard Generalized Markup Language (SGML;
ISO 8879-1986), SMDL files will probably normally be plain ascii text
files, which will easily transmit as e-mail or whatever.

------------------------------

Date: Wed, 24 Oct 90 19:05:32 BST
From: Nicholas Paul Carter <[e-mail]>
Subject: Optical musical score recognition
To: Music-Research@prg
Message-ID: <[e-mail]>

re: Music-Research Digest- Volume 5 : Issue  87
    and the messages concerning 'Optical musical score recognition'.

Contrary to Stephen Page's remark that "work in this field ... seems to have
died away", our work into 'automatic recognition of printed music' is continuing
here at the University of Surrey, as is that of several other research groups
around the world.
Anyone interested in this subject may wish to consult the following:-

"Computing in Musicology A Directory of Research", eds. Hewlett, W. and
Selfridge-Field, E., Center for Computer Assisted Research in the Humanities,
Menlo Park, CA., October 1989, 31-34.
"Automatic Recognition of Printed Music in the Context of Electronic
Publishing", Carter, N.P., PhD thesis, University of Surrey, February 1989.
"Acquisition, Representation, and Reconstruction of Printed Music by Computer:
A Review", Carter, N.P., Bacon, R.A. and Messenger, T., Computers and the
Humanities, 22(2), 1988, 117-136.
"Automatic Recognition of Music Notation", Carter, N.P. and Bacon, R.A.,
pre-Proceedings of the International Association for Pattern Recognition
Workshop on Syntactic and Structural Pattern Recognition, Murray Hill, NJ,
1990, 482.

Dr Nicholas Carter
Research Officer, Depts. of Physics and Music, University of Surrey, U.K.

------------------------------

Date: 22 Oct 90 14:57:03 GMT
From: "Mr. P. H. Smith" <[e-mail]>
Subject: Optical musical score recognition
To: music-research@prg
Message-ID: <[e-mail]>

In article <[e-mail]> [e-mail] (David Bainbridge) writes:
>  I am seeking information for my final year honours project at the
>department of Computer Science, Edinburgh University.  The project
>involves interpreting musical features (notes, sharp signs etc) from
>an image produced by an optical scanner.  Has anyone ever tackled a
>similar project, or know of any literature in this field ?


Alan Ruttenberg at the MIT Media Lab just showed me a demonstration of
his optical score reading stuff.  His system finds beams, note heads,
barlines, staff lines, accidentals, clefs, etc.  He's still working on
it and has yet to implement the midi transcription.  Anyway, you
should talk to him about it.

email: [e-mail]

Paul
[e-mail]

------------------------------

End of Music-Research Digest

--
* These words are mine alone, but I make no claim to original thought *