Measuring unpitched signal vs. pitched? Can this be done?

Recording Techniques, People Skills, Gear, Recording Spaces, Computers, and DIY

Moderators: drumsound, tomb

Post Reply
tallis
audio school
Posts: 2
Joined: Mon Sep 26, 2011 3:04 am

Measuring unpitched signal vs. pitched? Can this be done?

Post by tallis » Mon Sep 26, 2011 3:14 am

Hey there!

I'm wondering if you're aware of any means of measuring the relative amounts of pitched versus unpitched (i.e. noise) signal within a sample recording; it seems the sort of thing for which one could build a crude tool without too much trouble.

I am NOT looking for a measurement of baseline noise, like LEQ. Instead, I'm looking to measure things like how pitched a drum is, or how much white noise there is in a given distorted guitar tone.

A spectrograph can hint at these things visually; I'm going for a simpler set of data. Ideally, a ratio of pitch to "noise" averaged over a sample. Perhaps noise could be defined as something like multiple waves (within a FFT) of sufficient amplitude and insufficient spacing/constancy within a given range.

Thus, slow torch song with a rich voice and minimal drums might give you, say, 10:1, while a Throbbing Gristle record might give you something like 1:2.

Does anything like this exist?

User avatar
Nick Sevilla
speech impediment
Posts: 4825
Joined: Mon Mar 03, 2008 1:34 pm
Location: Los Angeles California USA
Contact:

Post by Nick Sevilla » Mon Sep 26, 2011 7:13 am

I can think of one such device :

Our brain.
Realizing vibratory excursions from a paper widget.

User avatar
A.David.MacKinnon
ghost haunting audio students
Posts: 3221
Joined: Wed May 07, 2003 5:57 am
Location: Toronto
Contact:

Post by A.David.MacKinnon » Mon Sep 26, 2011 7:14 am

I have no idea. I would guess that every natural sound has varying degrees of noise and that what we call tone is a reflection of the quality and quantity of that noise. You could probably argue that there is no such thing as a pure sound in the natural world. Even an solo vocal is affected by the room the singer is in, the size, shape and resonant qualities of the singers head, spit, phlegm and sinus conditions on the given day, atmospheric pressure, etc, etc, etc.

The bigger question for me is why you'd want to know and/or how would you apply that knowledge in any kind of useful way.

tallis
audio school
Posts: 2
Joined: Mon Sep 26, 2011 3:04 am

Post by tallis » Mon Sep 26, 2011 7:38 am

Sure, I can use my ears/brain. The reason I'd like such a tool is for analytic purposes. Suppose I'm writing an academic paper and I want to demonstrate postivistically timbral tendencies over a period of recording history, something like this could tell me a fair bit.

Thanks!

User avatar
Gregg Juke
cryogenically thawing
Posts: 3509
Joined: Fri Jun 11, 2010 10:35 pm
Location: Buffalo, NY, USA
Contact:

Post by Gregg Juke » Mon Sep 26, 2011 11:37 am

Sounds like a great topic for a paper!

But I also think you may have to think less scientifically/analytically (at least in terms of test-equipment verification), and put energy into some type of multi-factor grid/graph or chart that various recordings can be analyzed/broken down with. Criteria that will be fairly evident and clear to any and all listeners. This is where a project like yours could become difficult and get snagged, because these types of categories will always be subjective to a major degree, but you could start with definitions-- what engineers and audiologists (and maybe cognitive musicologists and ethnomusicologists) on the one hand consider "distorted," as compared to what musicians and music journalists called "distorted."

Good luck with that.

GJ

PS/Edit-- Of course, I just let my "music" bias show, as you could be approacching this from a purely technical standpoint, as in "How much distorion was in an 1896 Edison cylinder vs. a modern Mp3 or 180 gram audiophile vynil release?" Again, I don't think the gear you describe exists; good luck with that.

Maybe look into how/where figures for "THD" in a system actually come from? (I don't know how they measure that)...

User avatar
Snarl 12/8
ghost haunting audio students
Posts: 3326
Joined: Sat Dec 20, 2008 5:01 pm
Location: Right Cheer
Contact:

Post by Snarl 12/8 » Mon Sep 26, 2011 1:32 pm

You might want to check out some books on System Science. Bell Labs was a huge developer in that field and they developed one of the most successful and robust systems in the world (the phone system, when was the last time you didn't get a dial tone?)

Anyway, they worked out all kinds of equations related to distortion and signal because they needed to know what the working minimum signal clarity needed for two humans to communicate over a phone line was. It might not be exactly what you're looking for, but it might send you in the right direction and/or maybe their equations could be reworked from semantic information to musical.

If anyone has a patent on the device you're looking for, it'd be Bell Labs, I'd wager. There might be some piece of telephone testing equipment that does what you're looking for, it would probably be on an electronic signal not something coming out of a speaker.

ck

edit: http://cm.bell-labs.com/cm/ms/what/shan ... on1948.pdf
Carl Keil

Almost forgot: Please steal my drum tracks. and more.

The Scum
resurrected
Posts: 2426
Joined: Thu Jul 03, 2003 11:26 pm
Location: Denver, CO
Contact:

Post by The Scum » Mon Sep 26, 2011 8:56 pm

A friend of mine was showing me just such a meter, written in Max/MSP. It had a spectrogram, and along one edge was a rating of how noisy the input was. I seem to recall it being roughly instantaneous (averaged over a second, maybe). The thing was written in Max, so you could probably disassemble it, see how it works, and alter it to suit your needs.

You might look over some signal theory references. There are a bunch of different types of noise, with specific definitions. The signal theory folks categorize and measure appropriately.
"What fer?"
"Cat fur, to make kitten britches."

accordion squeezist
pushin' record
Posts: 218
Joined: Mon Jan 25, 2010 7:34 am
Location: Spring Grove, Pennsylvania

Post by accordion squeezist » Tue Sep 27, 2011 7:22 pm

you may be able to contact Wendy Carlos from her website, I think she knows this shit.

unchartedthickets
ass engineer
Posts: 45
Joined: Mon Feb 08, 2010 4:58 am

Post by unchartedthickets » Tue Sep 27, 2011 8:04 pm


Cyan421
takin' a dinner break
Posts: 192
Joined: Sat Nov 05, 2005 3:56 pm
Location: Idaho (On The Causeway to Neverwhere)

Post by Cyan421 » Wed Sep 28, 2011 12:40 am

Have people done double blind studies to try to figure out this sort of thing. That is if music has become more distorted or the years. Just having people listen to music and answer some multiple choice questions about it being negative, positive, loud, what have you. Interviews with 1000 music teachers would say yes. Many many engineers are desperately trying to add distortion to their mixes/recordings as a result of digital recording. Mixing stems to tape and back to digital, "tube" pre-amps, fake vinyl sounds. That is two unequivocal yeses from different sides of the fence.
"What a wonerful smell you've discovered"

User avatar
Gregg Juke
cryogenically thawing
Posts: 3509
Joined: Fri Jun 11, 2010 10:35 pm
Location: Buffalo, NY, USA
Contact:

Post by Gregg Juke » Wed Sep 28, 2011 8:43 am

True. But that is still pretty subjective. I think the OP is looking for a way to codify objective data of some kind.

GJ

Cyan421
takin' a dinner break
Posts: 192
Joined: Sat Nov 05, 2005 3:56 pm
Location: Idaho (On The Causeway to Neverwhere)

Post by Cyan421 » Wed Sep 28, 2011 11:30 pm

I'm saying I agree with Mr. Nick Sevilla, that the human brain is capable of recognizing how distorted something is. And with a large enough sample size it sorta is objective data, right?
"What a wonerful smell you've discovered"

User avatar
Gregg Juke
cryogenically thawing
Posts: 3509
Joined: Fri Jun 11, 2010 10:35 pm
Location: Buffalo, NY, USA
Contact:

Post by Gregg Juke » Thu Sep 29, 2011 4:14 am

>>>>And with a large enough sample size it sorta is objective data, right?<<<<

Yes! And that's what I was getting at too; I think some kind of grid, graph, or multi-response questionairre would be another way, if not the way, to go.

GJ

Post Reply

Who is online

Users browsing this forum: No registered users and 11 guests