Music File Compressed 1,000 Times Smaller than MP3

[Ichingcarpenter]

Researchers at the University of Rochester have digitally reproduced music in a file nearly 1,000 times smaller than a regular MP3 file.

The music, a 20-second clarinet solo, is encoded in less than a single kilobyte, and is made possible by two innovations: recreating in a computer both the real-world physics of a clarinet and the physics of a clarinet player.

The achievement, announced today at the International Conference on Acoustics Speech and Signal Processing held in Las Vegas, is not yet a flawless reproduction of an original performance, but the researchers say it's getting close.

"This is essentially a human-scale system of reproducing music," says Mark Bocko, professor of electrical and computer engineering and co-creator of the technology. "Humans can manipulate their tongue, breath, and fingers only so fast, so in theory we shouldn't really have to measure the music many thousands of times a second like we do on a CD. As a result, I think we may have found the absolute least amount of data needed to reproduce a piece of music."This research is funded by the National Science Foundation>>>>snip


Sound file using Bocko suppression wave file
http://www.rochester.edu/news/audio/sound4.wav

Human performance wave file
http://www.rochester.edu/news/audio/sound2_160mp3.wav

http://www.rochester.edu/news/show.php?id=3136


This is pretty important news on many levels

[zerox]

1. compression or synthesis?

This sounds more like music synthesizing than actual compression, which has been done for awhile. MIDI files are also very small, they just don't sound all that much like real performers.

[bullimiami]

2. that was my impression too. more like reproduction than recording.

[napoleon_in_rags]

5. see post #4 nt

[lurky]

3. So, to compress a symphony recording I have to

model each performer on every instrument, along with room acoustics, microphone placement, audience/ambient noise, and anything else I can thing up? I wouldn't call it compression, unless you call a midi file compression. Sounds more like synthesis or modeling, as someone else pointed out. And you will never have enough information to reproduce a recording in a satisfactory way. Not to mention all the processing you would have to do on your ipod to generate these sounds!

[napoleon_in_rags]

4. Is there a difference?

Now that I think about it, that's really interesting. The thing about compression is that its "complete", or anything you put in it can be compressed and sound decently like the original to the human ear. Does there exist some "physical system" for any sound you put it, and can it always be found? If so, synthesis could be compression.

P.S. oops, meant as a response to post #1

[lurky]

6. I guess technically, sheet music is compression.

:D

Actual "compression" techniques just find ways to store sound information in a smaller package, but don't have to know anything about how the sound was created. In some cases, like with mp3's, some of the information is lost, resulting in an imperfect reproduction, and therefore lower "quality" sound. What they describe is a much different technique than just mathematically approximating a wave form. They are trying to approximate a human musician. I think they are underselling the work by describing it as just compression.

[napoleon_in_rags]

7. Man, I just love this kind of stuff.

Even thinking about what you said about sheet music being compression is enough to get mental gears going. I remember hearing that any wave can be expressed as a sum of sines.

So a sheet music of sines could express any wave no matter how weird with infinite pitches, but sheet music doesn't have infinite pitches, so it would be limited. I wonder what it could express.

Speaking more generally to what you said, I think you are right. They are underselling it . But in a way, thinking of it as compression rather than synthesis really does open up more interesting doors. Because the really interesting problem is to take some strange music and try to come up with a viable physical system that could have produced it. That seems to have applications which stretch way beyond mere compression.

[Tesha]

8. You understand that...

> So a sheet music of sines could express any wave no matter
> how weird with infinite pitches, but sheet music doesn't
> have infinite pitches, so it would be limited. I wonder
> what it could express.

You understand that that's pretty-much how MP3 and AAC
do their compression today, right? They do it primarily
in the frequency domain rather than in the time domain.

Tesha

[Occulus]

10. A quibble:

Sheet music does indeed have infinite pitches. There is no hard limit that I'm aware of as to how many ledger lines above or below the staff can exist on a 'page'. They're representations of numbers, and as such can go on forever, in both the ultra- and infrasound directions.

Note that you can compress the notation of such extreme high and low pitches even further by using markings which act as placeholders for additional information, such as shifting by an octave. An example would be writing something marked with 8va, which would cause a musical passage to be played one octave higher than notated. In this way, one can eliminate the ledger lines- the extraneous information- and still notate the music so that it is properly performed.

Another example would be writing a glissando marking instead of each individual note.

I've never really thought of sheet music as a form of compression, but it's a perfectly apt description.

[gtar100]

9. This is probably more useful for talk than music

Which can translate into lower bandwidth requirements for digital voice transmission.

[arendt]

11. In speech, its called Linear Predictive Coding. Been around for thirty years...

http://en.wikipedia.org/wiki/Texas_Instruments_LPC_Speech_Chips

arendt

[BlooInBloo]

12. What's the big deal? The theorems of mathematics still stand...

.... The information-theoretic tradeoff still stands: lose accuracy to the original, or lose compressibility.

I could make an mp3 that's 1000 times smaller than a give mp3 too - just increase the lossiness. That is the article doesn't quantitatively discuss the *quality* of the "new" compression. They also don't discuss the processing requirements - it's cheating if the data file just says "go", and the player does all of the reproducing work.

There's nothing in the link to conclude much of anything.

[hunter]

13. The ultimate compression would probably be something that left a memory...

... identical to the memory that would have been left by the original experience.

This takes us to that scary place where the Universe might simply be a simulation running on some big computer somewhere...

Why program all the intricacies of the clarinet and it's playing, when you can simply keep track of it as less involved memories of what was purported to be occurring? Maybe all the stars in the sky are as fake as the ray traced image of a movie special effect. If you are looking at the stars with your eyes, maybe the image is compressed down to the wavelengths eyes can see. When you are looking at the stars with a sophisticated astronomical instrument, maybe the image is upgraded on the fly to be consistent within the properties of that instrument. Maybe the moon isn't actually modeled as a physical body unless someone happens to be walking on it. Heh. If a tree falls...

This compression technology is an interesting idea, somewhere on the road between midi and general purpose compression algorithms like ogg or mp3. In fact the "decoders" of ogg or mp3 systems are sometimes described as sophisticated musical instruments driven by a stream of numerical values. An mp3 or ogg decoder is much like a very complex version of a player piano.