Topics

erb, mel, parameters

Wim van der Meer <wimmeer@...>
 

What a great program Paul has made! I am a musicologist and
after fiddling a bit with the parameters in the "to pitch ..." dialog I
got exactly what I wanted. By relocating the pitch (multiply by the
inverse of a measured tonic) you can have your tonic at 0
semitones. Maybe cents would be more universal for a
musicologist, as many scales in the world are not based on a
twelve sem,itone system.
Can anyone inform me what erb and mel stands for (or where
can I read about them). And is there more information about the
analysis parameters. Fiddling is great, but to have some idea of
what I'm fiddling with might be nice.

Paul Boersma <paul.boersma@...>
 

Wim van der Meer wrote:
Maybe cents would be more universal for a
musicologist, as many scales in the world are not based on a
twelve sem,itone system.
neither are they based on a 1200-cents system...
But you just multiply by 100, of course.

Can anyone inform me what erb and mel stands for
they are psychophysical measures of frequency.
The Mel scale is supposed to be natural for perceived pitch,
since it is based on reported equal pitch distances.
The ERB (Equivalent Rectangular Bandwidth) scale is one of the scales
that linearizes the basilar membrane (like the Bark scale),
and is based on the maximum width of a band-limited white noise
of a given intensity that is reported as being perceived as equally
loud as a pure sine wave at the centre of this band (it is close to,
but not identical to, the critical bandwidth).
See any work on psychoacoustics (perhaps entitled "Hearing"
or "Audition"). Phoneticians could consult e.g. James Flanagan
(1965, 1972) "Speech analysis, synthesis and perception"
or Brian Moore's article in the Handbook of Phonetic Sciences.

And is there more information about the
analysis parameters. Fiddling is great, but to have some idea of
what I'm fiddling with might be nice.
the manual (click "Help" instead of "OK" in the "To Pitch..." dialog)
contains, besides relevant information about these parameters,
a pointer to the article (downloadable from my home page)
that describes the algorithm in detail.
--

Paul Boersma
Institute of Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam, The Netherlands
http://www.fon.hum.uva.nl/paul/
phone +31-20-5252385