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

Re: normalisation for spectral levels


Rather than getting hung up on formant-peak amplitude differences, it would
be more appropriate to look for changes in formant bandwidth and
peak-to-adjacent-trough ratio. Another way would be to see to what extent
the different exemplars diverge from either a generic neutral vowel or from
formants generated by a uniform tube, as suggested by Rene' Carre' (ask him
for reference, carre@tsi.enst.fr).

Pierre Divenyi

At 01:46 PM 9/24/2003 +0100, Christel de Bruijn wrote:
Dear list,

I apologise if the following question appears to be trivial.  I would be
very grateful for an answer though.

I would like to compare formant amplitude levels (in dB, measured from a
spectrum) of a sustained vowel before and after a reading task. However, I
adjusted the recording level for each vowel, in order to optimise SNR.

Therefore, it does not make sense to compare absolute dB levels.

I was wondering if I could do a simple normalisation for each vowel by
substracting its RMS value (after converting it to a dB value) from its
formant level.

Could I then compare the formant levels, or is this too simple?


Christel de Bruijn


Christel de Bruijn
Department of Human Communication Sciences
University of Sheffield, UK