There are 10 messages totalling 825 lines in this issue.
Topics of the day:
1. envelope extraction (2)
2. information transmission analyses (3)
3. job advert: Investigator Scientist at MRC-IHR, Glasgow
4. analysis of paired comparison data
5. early history of Auditory Neuroscience (2)
6. Fwd: early history of Auditory Neuroscience
----------------------------------------------------------------------
Date: Tue, 24 Feb 2009 09:26:19 +0000
From: Roy Patterson <rdp1@xxxxxxxxx>
Subject: Re: envelope extraction
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Hello Ben,<br>
<p class="MsoNormal"><br>
You are right to question auditory models with transduction filters
whose cutoff is below about 1000 Hz. They are often just trying to
explain amplitude modulation detection without concern for other
aspects of auditory processing.<br>
</p>
<p class="MsoNormal">There is a discussion of the different methods of
extracting
envelopes in Yost et al. 1998, reference below. It concludes that if
you want
to be able to explain the pitch shift of the residue, there is only one
of the
methods that can be used to represent auditory processing, and that is
half-wave rectification followed by lowpass filtering. The filter needs
to have
a relatively high cutoff frequency, 800-1200 Hz and beyond the cutoff,
the
filter should fall about 24 dB per octave.<o:p> </o:p></p>
<p class="MsoNormal">Regards Roy P<br>
<br>
Yost, W.A., <strong>Patterson, R.D. </strong>and Sheft, S. (<strong>1998</strong>).
The role of the envelope in processing iterated rippled noise. <em>J.
Acoust.
Soc. Am.</em> <strong>104</strong> 2349-2361.</p>
<br>
<br>
<br>
Hornsby, Benjamin Wade Young wrote:
<blockquote cite="mid:20090223170018.3737951E2@xxxxxxxxxxxxxxxxxxxxxxx"
type="cite">
<meta http-equiv="Context-Type" content="text/html; charset=us-ascii">
<div>
<p> <span>Hi all, </span> </p>
<p> <span>I hope the answer to this question is not too obvious
but,…
We have recently been talking about auditory processing models, in
specific, temporal
processing models, many of which incorporate an envelope extraction
stage. To
do this I’ve seen in many cases the use of half wave rectification
followed by a low pass filter. The cutoff for this filter is generally
a pretty
low frequency, say around 50 Hz or so. A colleague suggested that the
actual cutoff
frequency should be based on the assumed time constant of the system or
subsystem being evaluated. I’ve been trying to determine the
physiologic rational
for such a low frequency filter (slow time constant) in models of
auditory
processing. Neural processing limitations like the refractory period of
the
neuron are much faster than this. Anyone have a reference or two that
might
help explain the physiologic rational for this low frequency filter
cutoff? </span> </p>
<p> <span> </span> </p>
<p> <span>Thanks much, </span> </p>
<p> <span> </span> </p>
<p> <span>Ben </span> </p>
<p> <span> </span> </p>
<p> <span> </span> </p>
</div>
</blockquote>
<br>
<br>
<pre class="moz-signature" cols="60">--
* ** *** * ** *** * ** *** * ** *** * ** *** *
Roy D. Patterson
Centre for the Neural Basis of Hearing
Department of Physiology, Development and Neuroscience
University of Cambridge
Downing Street, Cambridge, CB2 3EG
<a class="moz-txt-link-freetext" href="http://www.pdn.cam.ac.uk/cnbh/";>http://www.pdn.cam.ac.uk/cnbh/</a>
phone: +44 (1223) 333819 office
fax: +44 (1223) 333840 department
email <a class="moz-txt-link-abbreviated" href="mailto:rdp1@xxxxxxxxx";>rdp1@xxxxxxxxx</a>
</pre>
</body>
</html>
------------------------------
Date: Tue, 24 Feb 2009 09:29:24 +0000
From: caroline jacquier <caro_jacquier@xxxxxxxx>
Subject: information transmission analyses
--0-282690217-1235467764=:86427
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Dear members,=0A=0AIwould like to analyse phonetic feature transmission (es=
pecially voicing and place of articulation of french plosive consonants) on=
the basis of individual confusion=C2=A0matrices. My study is about compres=
sed speech perception in dyslexic adults. =0AI have already downloaded the =
software for information transfer and SINFA analysis (Stuart Rosen ) and I =
have run the analyses (on=C2=A032 subjects).=0AMy first question is: which =
value is the more relevant in the report (TRANS, TRANS/INPUT or TRANS/TI) ?=
=0AMy second question is about individual confusion matrices: Do I have to =
do one analyse per subject ?=C2=A0=0A=C2=A0=0AI hope that someone could hel=
p me.=0AThank you very much.=0ABest regards,=0A=C2=A0=0ACaroline=0A=0A=0A =
--0-282690217-1235467764=:86427
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable
<html><head><style type=3D"text/css"><!-- DIV {margin:0px;} --></style></he=
ad><body><div style=3D"font-family:times new roman, new york, times, serif;=
font-size:12pt"><DIV>Dear members,</DIV>=0A<DIV> </DIV>=0A<DIV>I<FONT =
face=3D"times new roman, new york, times, serif"> would like to analyse pho=
netic feature transmission (especially voicing and place of articulation of=
french plosive consonants) on the basis of individual confusion matri=
ces. My study is about compressed speech perception in dyslexic adults. </F=
ONT></DIV>=0A<DIV><FONT face=3D"times new roman, new york, times, serif">I =
have already downloaded the s<SPAN class=3D375500014-17022009>oftware for i=
nformation transfer and SINFA analysis (Stuart Rosen ) and I have run the a=
nalyses (on 32 subjects).</SPAN></FONT></DIV>=0A<DIV><SPAN class=3D375=
500014-17022009><FONT face=3D"times new roman, new york, times, serif">My f=
irst question is: which value is the more relevant in the report (TRANS, TR=
ANS/INPUT or TRANS/TI) ?</FONT></SPAN></DIV>=0A<DIV><SPAN class=3D375500014=
-17022009><FONT face=3D"times new roman, new york, times, serif">My second =
question is about individual confusion matrices: Do I have to do one analys=
e per subject ? </FONT></SPAN></DIV>=0A<DIV><SPAN class=3D375500014-17=
022009></SPAN> </DIV>=0A<DIV><SPAN class=3D375500014-17022009>I hope t=
hat someone could help me.</SPAN></DIV>=0A<DIV><SPAN class=3D375500014-1702=
2009>Thank you very much.</SPAN></DIV>=0A<DIV><SPAN class=3D375500014-17022=
009>Best regards,</SPAN></DIV>=0A<DIV><SPAN class=3D375500014-17022009></SP=
AN> </DIV>=0A<DIV><SPAN class=3D375500014-17022009>Caroline</SPAN></DI=
V>=0A<DIV><FONT face=3DArial size=3D2><SPAN class=3D375500014-17022009></SP=
AN></FONT> </DIV>=0A<DIV><BR></DIV></div><br>=0A=0A=0A=0A </body>=
</html>
--0-282690217-1235467764=:86427--
------------------------------
Date: Tue, 24 Feb 2009 11:51:04 +0000
From: Stuart Rosen <stuart@xxxxxxxxxxxxxxxxxxx>
Subject: Re: information transmission analyses
I highly recommend you read the classic book by Attneave in order to get
to grips with information transfer measures generally:
Attneave F. Applications of information theory to psychology: a summary
of basic concepts, methods, and results. New York: Holt, Rinehart, and
Winston, 1959. 120 p. [University of Oregon, Eugene, OR]
In my experience %TRANS (=TRANS/INPUT) has proved to be the most useful
measure but TRANS/TI can give you an idea about how important any
particular feature is in overall performance. TRANS itself is probably
less useful because the amount of information in each feature differs
considerably.
How you do the analyses depends on what questions you are trying to
answer. Analysing individual matrices and then doing standard
statistical tests is one way to avoid the difficulties of doing
statistical analyses on a small number of matrices (which require
something like log-linear analyses). On the other hand, looking at group
matrices can make it easier to spot trends.
Bon chance!
Yours - Stuart