Sound Textures

Auditory scientists have spent much time studying harmonic waves. But clearly most sounds that we hear are more complicated than simple harmonic sounds.

The sound of rain, or crowds, or even copy machines all have a distinct temporal pattern which is best described as a sound texture. This class of sounds is often overlooked, as most researchers focus on more salient sounds, like speech or music; for them, a sound texture is simply noise, that needs to be factored out.

Between 1993 and 1995, I studied sound textures at the MIT Media Lab. First I explored sound textures from a human perception point of view, performing similarity and grouping experiments, and looking at the features used by subjects to compare sound textures.

Second, from a machine point of view, I developed a restricted model of sound textures as two-level phenomena: simple sound elements called atoms form the low level, and the distribution and arrangement of atoms form the high level. After extracting the sound atoms from a texture, my system used a cluster-based probability model to characterize the high level of sound textures. The model is then used to classify and resynthesize textures. Finally, I draw parallels with the perceptual features of sound textures explored earlier, and with visual textures.

_____________________

Here is my thesis (5MB PDF), and here is a shorter book chapter (1.1MB PDF) on resynthesizing Sound Textures, which actually provides a better read than the whole thesis.

_____________________

nsa(at)media(dot)mit(dot)edu




Related Work

Over time, several researchers have contacted me about Sound Textures. I apologize if I have not always responded in a timely fashion. In an effort to connect this community, I have copied below some of the messages I received over the last few years. If there is enough interest, I will setup a mailing list on the topic. Also, if at least one of you wants to contribute samples of sound textures (WAV, MP3 or OGG files), I will create a repository for everyone to access.

Keep in touch, and send me URLs of related work!


-----Original Message-----
From: Maria Neimark Geffen [mailto:mgeffen(at)rockefeller(dot)edu] 
Sent: Wednesday, March 21, 2007 1:21 PM
Subject: sound texture analysis


_____________________



-----Original Message-----
From: Gerda Strobl [mailto:gerda.strobl(at)gmail(dot)com] 
Sent: Tuesday, February 20, 2007 12:43 PM
Subject: a new work on textures - parametric sound texture generator

Dear Nicolas,

in January I finished my master thesis on sound texture generation 
(analysis/resynthesis)-perhaps you find some interesting ideas:

http://iem.at/projekte/dsp/parametric/project_view

Of course there is a reference to your former work. Actually I only 
had the article from the book, because somehow- I did not reach that 
web for a long time.

Hope you enjoy reading some part of it!

Best regards, 
Gerda


_____________________



-----Original Message-----
From: raja ghozi [mailto:rjghozi(at)yahoo(dot)com] 
Sent: Thursday, December 21, 2006 4:27 AM
Subject: Audio texture

Dear Nicholas,
 
One of my students has already contacted you regarding your work on
textured sounds. We are very interested in this class of signals and
we believe that there is so much to be explored, especially with 
reference to visual texture.
 
We would like to sollicit your help in getting samples (actually a 
variety) of audio textures.

It would also be nice if you ould point us to individuals working on 
this topic.
 
Best regards,
Raja Ghozi, PhD
 
 

_____________________


Date: Sun, 5 Nov 2006 12:08:40 +0100 (CET)
From: oussema Lengliz [mailto:oussema_lengliz(at)yahoo(dot)fr]
Subject: a student working on Sound Textures
Content-Transfer-Encoding: 8bit

hello Mr Nicolas,

  I'm a student at l'ENIT (Ecole Nationale des Ingénieurs de Tunis),
TUNISIA, second level in the Telecom Field.

  I'm now working, with my friend on the theme of sound texture.
We have submitted to a proposition of the Laboratory U2S (Unité S
ignaux & Systèmes)  with Mms GHOZI Raja and LARBI Sonia to work 
on analyzing and clasifing the sound textures.

  So, we are looking for a further help from you.

  thank you,
  Oussema Lengliz
  2ATEL


_____________________


From: Sarah Sussman [mailto:ssussman(at)stanford(dot)edu] 
Sent: Wednesday, January 24, 2007 11:42 AM
Subject: sound textures


http://www.guardian.co.uk/science/story/0,,1997234,00.html?gusrc=rss&feed=1

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

Sarah Sussman
Curator, French and Italian Collections
Stanford University Libraries
Green Library West (Bing) 320A
Stanford CA 94305-6004

_____________________


From: "Donovan irish" [mailto:spandex911(at)hotmail(dot)com]
Date: Wed, 13 Sep 2006 04:36:22 +0000

i stumbled across your thesis on sound textures, or at least, a breif 
synopsis of your research,,,was wondering if there is anywhere i can read 
more of this research, its very interesting to me.
thanks
donovan

_____________________


Date: Fri, 24 Mar 2006 17:08:32 +0100
From: Jean-Julien Aucouturier [OBSOLETE: jj(at)csl(dot)sony(dot)fr]
Organization: SONY CSL
Subject: Re: IJCAI 95 paper

oops looks like the much-vaunted picture won't paste in the email.

It shows 4 different curves of "potential information" againt time, for 
music and speech on the one hand (steep, linear)
and for textures and noise on the other hand (logarithmic)
It is labelled by Peltonen in his reproduced version
'Constant long-term characteristics of sound textures and noise'
Hope this helps clarifying.

Thanks in advance,
Jean-Julien


Jean-Julien Aucouturier wrote:

> Hello,
>
> I'm trying to get a copy of your IJCAI 95 paper on sound textures, or 
> any other published work which
> gives details about the figure embedded below:
>
>
>
>
> as found in a report by Peltonen from Tampere, Finland (Vesa Peltonen
> COMPUTATIONAL AUDITORY SCENE RECOGNITION, 2001)
> and which is supposingly from your paper.
>
> The reason I'm interested is that I'm writing a paper on a technique 
> to establish the same kind of information rates
> for pattern recognition algorithms, compared for polyphonic music and 
> urban sound textures.
> I'd like to have this properly quoted if relevant.
>
> Thanks !
> Jean-Julien

-- 
Jean-Julien Aucouturier, Assistant Researcher
OBSOLETE: http://www.csl.sony.fr/~jj
SONY CSL Paris 6, rue Amyot 75005 Paris               

_____________________


Subject: Master thesis
Date: Sun, 31 Jul 2005 05:17:49 +0200
From: "Stefania Serafin " 

Dear Nicolas,

   my name is Stefania Serafin and I am assistant professor in Medialogy 
at Aalborg University in Copenhagen. I work on sound synthesis mostly by 
physical models. Recently I became interested in audio textures.

I was wondering if I could have a copy of your master thesis, since I 
cannot find it online.

Thanks in advance

Stefania Serafin


www.media.aau.dk/people/sts

_____________________


Subject: Texture Perception
Date: Wed, 16 Nov 2005 10:33:08 -0000
From: "Sukhbinder Kumar" [mailto:Sukhbinder(dot)Kumar(at)newcastle(dot)ac(dot)uk]

Dear Nicholas
 
I am a post doctorate research associate working in the auditory 
group of school of neurolobiology at university of newcastle (UK). 
I am interested in sound texture perception but unfortunately I 
have not been able to found references to literature on this topic. 
I could learn from your website (at MIT) that you have worked on 
texture perception in the past. I would be highly grateful to you 
if you could provide refernces to your work (ans also, if possible 
any other key work that you are aware of). 
 
best regards
Sukhbinder Kumar
Post doctoral Research Associate
Auditory System Group
School of Neurology, Neurobiology and Psychiatry,
Medical School, Framlington Place
University of Newcastle
Newcastle upon Tyne (UK)
NE2 4HH

_____________________