Klarinet Archive - Posting 000240.txt from 2002/12

From: John Dablin <jdablin@-----.com>
Subj: Re: [kl] I can name that tune in nine notes...
Date: Tue, 10 Dec 2002 15:09:09 -0500

The idea that a tune can be identified by simply by the up and down pattern
of the melody is not an original one, although it's possible that they made
the discovery independently. One of my prize possessions is "The Directory
of Tunes and Musical Themes" by Denys Parsons, (published in 1975 by Spencer
Brown and co, 17 Halifax Road, Cambridge, England, ISBN 0 904747 00 X). To
identify a tune you write down an asterisk for the first note, and then U, R
or D for the successive notes depending on whether the note goes Up, Remains
the same, or goes Down compared to the one before it. Then you look it up
alphabetically in the book. So, if I write down *UURDD DUUUR DDDUD, it
tells me that it's "Mozart clarinet concerto in A K622 2m" Amazingly, this
is a unique entry, there are very few duplications, and when there are it's
usually obvious from the style of the piece which one it is.

It must have been a herculean task to produce the book. There are 184 pages
of entries for classical themes and 67 of popular tunes. With 55 or 56
entries on each page there must be over 14000 entries.

A note on the jacket says that "Denys Parsons is the son of Alan Parsons, the
dramatic critic, and actress Viola Tree. He first came into prominence with
an anthology of misprints called "It Must be True, it was All in the Papers".
IIlustrated by Ronald Searle, it was an immediate success, and was followed
by others. His next venture was into more serious literature, in the shape
of the now classic guide, "What's Where in London". Considered an
indispensable work of reference, it has been published in many forms. The
present "Directory of Tunes and Musical Themes" took Mr Parsons five years to
complete, and embodies what is perhaps the most original and useful idea ever
to be devised for indexing melodies. Mr Parsons is now Press Officer to the
new British Library, formerly the library of the British Museum"

I've never heard of the book since I bought my copy. It would be invaluable
if it could be updated from time to time.

John Dablin

On Tuesday 10 December 2002 19:38, you wrote:
> On TV yesterday, there was a short article about humming. They
> demonstrated someone humming into a computer to try to figure out what song
> they were humming. The computer then searched a database and dug out the
> tune. I did a web search and found the paper.
>
> http://www.cs.cornell.edu/Info/Faculty/bsmith/query-by-humming.html
>
> The paper is about a scheme for being able to figure out what a song is on
> the basis of being able to listen to someone hum a snatch of the song.
>
> My thought was. "Some people hum really badly. How well could they do
> this?"
>
> They came up with a scheme to counteract this...
>
> It is a really cool idea----You do not have to hum well, they use a three
> letter alphabet - same as previous note, up from previous note, or down.
>
> > For example, the introductory theme Beethoven's 5th Symphony would be
> >converted into the sequence: - S S D U S S D (the first note is ignored
> > as it has no previous pitch).
>
> Anyway, I can imagine "Google song search".
>
> Great thing for a bar.
>
> The only issue is that they were using a 183 song database. In order to be
> "real" this would have to contain thousands of songs. They would probably
> also have to figure out a way to plug a CD or an MP3 into the analyzer and
> push out a representation rather than just using midi files. This was done
> in the day of the P-90. I suspect that there would be a lot more cheap
> computing power available today :-). They also talked about MP4 - which
> they described as a "postscript for sound" - instead of a representation of
> the waveform, it was instructions on how to reconstruct the waveform, and
> supposedly it is 1% the size of an MP3 (http://sound.media.mit.edu/mpeg4/)
> and maybe there would be a way to use information kept there for this sort
> of search.
>
> Anyway, they found that nine notes were generally enough to determine which
> song it was.

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