[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Loudness of Sounds

Thanks so much to all those who replied. I forgot to mention in my
original mail that we are primarily interested in MRI activations
caused due to the nature of the stimuli and not due to loudness.
However, we want to control for the loudness, so that it does not
act as an additonal cue.

I am appending a text file with the responses I received.



Date: Tue, 17 Apr 2001 13:21:56 -0500 (EST)
From: Brian Gygi <bgygi@indiana.edu>
To: Fatima Husain <fhusain@cns.bu.edu>
Subject: Re: Loudness of sounds

Hi Fatima,

In our lab we use environmental sounds (naturally occurring non-speech
stimuli) which tend to be broadband and to have complex temporal
envelopes, with large gaps of silences and strong transients.  We have
used two methods:  one is to equate the peak RMS in a 100 ms window for
the sounds.  This ensures that the loudest bursts are the same amplitude,
but it also means that steady-state sounds tend to be louder overall.
Another method we used (which we prefer) is to equate the RMS after
subtracting silences of greater than 50 ms.  This is called the
'pause-corrected RMS' and is recommended for speech in Ludvigsen, C.
(1992).  "Comparison of certain measures of speech and noise level,"
Scand. Audiol. 21, 23-29.

I hope this helps,

Brian Gygi

Date: Tue, 17 Apr 2001 14:43:05 -0400
From: Jont Allen <jba@research.att.com>
Subject: Re: Loudness of sounds

Fatima Husain wrote:

> We keep running into this problem in my lab and I am wondering if you could
> help. We often have stimuli sets that includes both speech and non-speech
> stimuli. How do we equalize their loudness?

One solution to this problem (I am reading between the lines here), is
to rove the level to remove any loudness cues. Instead of equalizing the loudness,
remove it as a reliable cue. Of course this is not answering the question
you asked, but may help with your ultimate goal.

> The first step is to see
> that they all have the same RMS power. But, the different stimuli being
> broadband and narrowband,the stimuli may still seem to have differing loudness.
> Is this enough?
> If not, what do we do as the next step? Do we measure output at the end of
> the headphones and equalize then? This is not always possible, esp. for
> headphones which are already part of the fMRI scanner...or we are piping
> sounds through a convoluted system. I am asking primarily with reference
> to fMRI studies.
> Any info if appreciated. Thanks! I will post a summary of responses on
> the list.

If you really want the loudness to be the same, you may need to do an experiment
before hand, that sets them equal. This would be tedious of course.

> --fatima
> fhusain@cns.bu.edu
> fthusain@helix.nih.gov



From: "Donald D. Greenwood" <ddg@neteze.com>
Date: Tue, 17 Apr 2001 12:12:00 -0700
Subject: Re: Loudness of sounds

Various respondents will write you that since power is a property of the
stimulus whereas loudness is an attribute of sensation (not a parameter of
the stimulus), you cannot expect sounds equated on one basis to be equal
on the other.  However, few will point you to the following reference,
which should be included in your reading.

Cacace, A.T. and Margolis, R.H.  (1983)   On the loudness of complex
stimuli and its relationship to cochlear excitation.  J. Acoust. Soc. Am.
78, 1568-1573.

I hope that you find it useful.

D. D. Greenwood

Date: Wed, 18 Apr 2001 09:06:23 +0100 (BST)
From: John Culling <jfc@brentwood.psyc.cf.ac.uk>
Subject: Re: Loudness of sounds

At the very least you should use an A-weighted measure of rms.
Whether you go beyond that to using a loudness model depends on
how inportant loudness matching is for you.

From: "Ward Drennan" <ward@ihr.gla.ac.uk>
To: "Fatima Husain" <fhusain@cns.bu.edu>
Subject: Re:      Loudness of sounds

We present a speech signal and noise in a dome-shaped chamber to measure
localization ability. To equalize the levels we created waves with the
same RMS and then equalized the levels using a sound level meter at the
center of the dome. Effectively, this is the same as measuring the level
at the end of the headphone and equalizing them. Could you place a Kemar
manikin in the fMRI and record through it, or hook up a sound level
meter through it?

Loudness is a subjective judgement. If you wanted to measure that,
could you place the listeners where you want them, play the sounds and
give them the ability to adjust the levels themselves until the loudness
is matched?

It would seem to me matching levels through headphones with a Kemar
manikin is the best bet. Would that work for you?

Ward Drennan, Ph. D.
Hearing Scientist


From: jan schnupp <jan.schnupp@physiol.ox.ac.uk>
Subject: Re: Loudness of sounds

Dear Fatima,

there is no perfect way of equalizing loudness, or rather, the best way
will depend on the questions that you are asking. If your stimuli are very
different in bandwidth then equalizing the overall RMS energy can lead to
very different energies per active critical band in the auditory system. As
a consequence, the "weaker" narrowband sound could, for example, give you a
stronger MRI signal because it focusses it's energy on a few bands, rather
than dissipating it over many.
Loudness perception is quite involved. B. Moore's chapter on loudness in
"An Intro to the Psych of Hearing" is not a bad place to start reading up
about all the factors that come in to it. Bandwidth is not the only factor
that may have to be taken into account. If you are sometimes looking at
very broad bands you may have to do something like A weighting to take into
account the lower sensitivity at the edges of the auditory range. And
temporal factors may play a role if some of your stimuli are a lot more
"impulsive" than others. Whether you need to worry about these things at
all depends entirely on the exact question you are asking and you will have
to review your strategy for equalising loudness from one experiment to
If for you loudness is just a "nuisance factor" and all you are trying to
do is to ensure that any systematic loudness variation does not obscure
some other aspect which you are really interested in, then perhaps you
could get the subjects to adjust the intensity of the test stimuli before
the experiments so that they sound about equally loud, and then during the
experiment vary the intensity of each stimulus randomly around that
subjectively equalised setting. If the patterns you get are relatively
insensitive to the variation in intensity then you can argue that loudness
was not important.

Hope that helps,


Dr. Jan Schnupp
Oxford University, Laboratory of Physiology, Parks Road, Oxford OX1 3PT, U.K.
Tel (+44-1865) 272 513  Fax (+44-1865) 272 469

From: Larry Shotland <shotland@nidcd.nih.gov>
To: fthusain@helix.nih.gov
Subject: loudness


You might want to measure the bandwidth of your signals and adjust
loudness with this in mind, as increased bandwidth of otherwise equal
signals will equate to a higher sound pressure.

Larry Shotland, Ph.D.
Hearing Section, Neuro-Otology Branch
National Institute on Deafness and Other Communication Disorders
National Institutes of Health
9000 Rockville Pike
Building 10, Room 5C-306
Bethesda, MD 20892