Music symbol recgonition
Ichiro Fujinaga
recognition (by GO?) in this group . Is there an archive for this group?
Can someone kindly post a summary of that discussion?
Thank you.
--
Ichiro Fujinaga (i...@music.mcgill.ca)
Faculty of Music, McGill University
555 Sherbrooke W. Montreal (Quebec) CANADA H3A 1E3
JeanRenard Ward
To: comp.sys.pen, 1084
Subject: Rumored music symbol recognition for GO
I don't know anything about music symbol recogniton at GO:
arguably, this is one case where pen input with HWX would be MUCH
inferior to the use of a piano-type keyboard.
However, one of the earliest technical papers to use the term "gesture
recognition" dealt with gesture input for a simplified form of
standard music notation:
Buxton, W. Sniderman, R. Reeves, W, Patel, S.., and Baecker, R., "The
evolution of the SSP score editing tools", Computer Music Journal, Vol 3,
1979, pp 14-25.
Bill Buxton was previously at the European office of XEROR PARC, where he
did a lot of this sort of stuff. I met him back in 1986 or so when the company
I was with at the time (Pencept) did a series of presentations on
commercial HWX and gesture-based UIs at that years CHI conference.
-- Jean Renard Ward
jrw...@world.std.com
Ketil Albertsen,TIH
>>I've heard that there was a discussion about music symbol
>>recognition (by GO?) in this group . Is there an archive for this group?
>>Can someone kindly post a summary of that discussion?
>
>I don't know anything about music symbol recogniton at GO:
>arguably, this is one case where pen input with HWX would be MUCH
>inferior to the use of a piano-type keyboard.
A recent issue of PC Magazine (March 30? Someone has run away with
the copy!) had a short notice about a commercial product running
under MS-Windows. It was claimed to analyze an average page of music
in 5 *minutes*, from a scan file in eg. TIFF format (so the scanning
time was not included). When you talk about processing times in the
minutes range, you make sure to measure it on the fastest 66 Mhz 486
available, with at least 16 Mbyte RAM! My PC is a 25 Mhz 386, so I
could probably go out for dinner while the program was working.
So, my conclusion is similar: Manual input is probably much faster,
has a lower error rate and includes more information (eg. dynamics).
(I have tried Cakewalk Pro - it is unsatisfactory and I am making my
own now, but I would expect even Cakewalk Pro to beat any sheet music
reader by a fair amount. - I chose to use a mouse for data entry, not
a keyboard - much easier with 4-part SATB vocal music!)
Alejandro Rivero
>In article <C669C...@world.std.com>, jrw...@world.std.com (JeanRenard Ward) writes:
>
>>>I've heard that there was a discussion about music symbol
>>>recognition (by GO?) in this group . Is there an archive for this group?
>>>Can someone kindly post a summary of that discussion?
>>
Could the discussin be in Compuserve? Perhaps anyone could abstract it
here...
>>I don't know anything about music symbol recogniton at GO:
>>arguably, this is one case where pen input with HWX would be MUCH
>>inferior to the use of a piano-type keyboard.
What is HWX? Hand Writing geXtures?
Hey guys, It must be a lot easier than text, must it? Music is that
kind of things you write always over a paper pauted with five
horizontal lines, and the sing depends only of relative position
in the line... or thing has changed actually?
Gift me a SDK and I will send you a demo version in two days.
>
>A recent issue of PC Magazine (March 30? Someone has run away with
>the copy!) had a short notice about a commercial product running
>under MS-Windows. It was claimed to analyze an average page of music
>in 5 *minutes*, from a scan file in eg. TIFF format (so the scanning
>time was not included). When you talk about processing times in the
>minutes range, you make sure to measure it on the fastest 66 Mhz 486
>available, with at least 16 Mbyte RAM! My PC is a 25 Mhz 386, so I
>could probably go out for dinner while the program was working.
>
Here the problem is that you must differentiate the pentagram from the
rest of data. And you have not dinamical info of the writing, so you
must analize where the ties are really.
>So, my conclusion is similar: Manual input is probably much faster,
>has a lower error rate and includes more information (eg. dynamics).
>
Microphone input is surely much faster, it has Zero error rate with the
adequate equipment and includes all the information.
I think someone is missing the goal here.
ALejandro Rivero
Theoretical Physics
Zaragoza Univ, Spain
Dale Frye
> In article 167...@lumina.edb.tih.no, ke...@edb.tih.no (Ketil
Albertsen,TIH) writes:
> >In article <C669C...@world.std.com>, jrw...@world.std.com (JeanRenard
Ward) writes:
> >
> >>>I've heard that there was a discussion about music symbol
> >>>recognition (by GO?) in this group . Is there an archive for this
group?
> kind of things you write always over a paper pauted with five
> horizontal lines, and the sing depends only of relative position
> in the line... or thing has changed actually?
We had a student write a Pen Music program last summer. Turns out it's not
as easy as it looks. Recognition of the type of note (quarter, half, etc)
is not a problem. The big problem is determining where the note should go.
The problem is that the vertical spacing between note positions overlaps a
great deal. When people write notes they can easily draw big circles that
overlap more than one bar. The recognizer must determine where the note
goes. Center of gravity gets most right but not all. A alternative
solution would be to use a different guesture (instead of a circle) to
represent the base of a note.
It was fun writing a tune, pressing the play button and listening to what
you just wrote.
Dale Frye
Pen Lab
Washington University in St. Louis
Sharpened Software
>In article <C669C...@world.std.com>, jrw...@world.std.com (JeanRenard Ward) writes:
>
>>I don't know anything about music symbol recogniton at GO:
>>arguably, this is one case where pen input with HWX would be MUCH
>>inferior to the use of a piano-type keyboard.
>
>A recent issue of PC Magazine (March 30? Someone has run away with
>the copy!) had a short notice about a commercial product running
>under MS-Windows. It was claimed to analyze an average page of music
>in 5 *minutes*, from a scan file in eg. TIFF format (so the scanning
>So, my conclusion is similar: Manual input is probably much faster,
>has a lower error rate and includes more information (eg. dynamics).
>
>(I have tried Cakewalk Pro - it is unsatisfactory and I am making my
>own now, but I would expect even Cakewalk Pro to beat any sheet music
>reader by a fair amount. - I chose to use a mouse for data entry, not
>a keyboard - much easier with 4-part SATB vocal music!)
There is a big difference between OCR and HWX! and an every bigger difference
between mouse input and pen input.
Entering notes in handwritten manuscript (straight balls and flags) form
gives the computer much more information than in OCR, thus the problem should
be much simpler. In OCR, the computer has to find the staff and determine
its scale and position. In HWX, the computer hands you the staff, eliminating
some of the error.
In most of the DOS/Windows/Mac software, notes are entered in two steps:
Pick a note type from a palette of notes, and put the note on the staff.
On a pen machine, you can just write the notes in the correct positions,
drawing the flags and dots yourself. It should also be possible to use the
palette method. Since pen input is more favorable toward random access,
even this style should be faster than a mouse.
--
__________________________________________________________________________
Michael Libes Internet: sha...@halcyon.com
Sharpened Software Inc. Compuserve: 71551,3667
James A Landay
>I don't know anything about music symbol recogniton at GO:
>arguably, this is one case where pen input with HWX would be MUCH
>inferior to the use of a piano-type keyboard.
Dean Rubine did an interesting music application that used gestures
to enter notes. It was part of his thesis, "The Automatic Recognition
of Gestures", CMU-CS-91-202. You might send him email at:
dean....@cs.cmu.edu
if you have interests in this area.