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

Re: AUDITORY Digest - 17 Nov 2002 to 18 Nov 2002 (#2002-197)



Dear Kala,

Phone recog. can have a zero variance when the signal to noise ratio is
good. Most
errors are due to production, not perception. The interesting question is
HOW the errors in the confusion matrix vary as a function of the SNR,
for SNR less than 30.
Look the many papers on the articulation index for the detail. As an
introduction
look at:

http://auditorymodels.org/jba/PAPERS/ICASSP/WCA.ps

Back up on the URL, and pick a format.
Jont

btw, I live in mountainside NJ, <15 miles away from Newark, or Rutgers.

Automatic digest processor wrote:

There is one message totalling 26 lines in this issue.

Topics of the day:

 1. Categorical Perception

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

Date:    Mon, 18 Nov 2002 13:44:20 -0500
From:    Kala Lakshminarayanan <kala@PEGASUS.RUTGERS.EDU>
Subject: Categorical Perception

Hi Everyone

I was asked recently what the test-retest reliabilities of categorical
perception identification and discrimination paradigms of phoneme
recognition are and i realised i dint know. I dont know if anyone has
explicitly tested the reliability. Or is it simply a population effect
and reliability is not a relevant question?

Thanks
Kala

--
Kala Lakshminarayanan
The Infancy Studies Laboratory
Center for Molecular and Behavioral Neuroscience
197, University Ave,
Newark, NJ 07102

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

End of AUDITORY Digest - 17 Nov 2002 to 18 Nov 2002 (#2002-197)
***************************************************************

--
Jont B. Allen,     jba@auditorymodels.org;   908/654-1274voice; 908/789-9575 fax
382 Forest Hill Way
Mountainside NJ 07092
http://auditorymodels.org/jba

``A paradox is simply an error out of control''
  --E.T. Jaynes, Chapter 15 of http://bayes.wustl.edu/etj/prob/