|
Hi
Thank you very much for the response.
I am using Praat both for synthesis and analysis (I'd like to use Klatt for synthesis but we don't have it here). The pich is 115 Hz. The algorithm used for linear prediction coefficients is the Burg algorithm. The remaining formant values are 1160; 2500; 3500; 4500...up to F10. The bandwidths are 50 (F1) ,100 (F2), 150, 200, 300, 400, 500, 600, 700, 800 (is it OK?).
I have maked two vowel continua: the first one is a F1 continuum from 300 Hz to 700Hz (step size = 10 Hz) simply as synthesised by Praat. The second one is the first one adjusted according to the analysis results (I have modified each sound, one by one, using the command "modify: formula").
So the analysis is the main source of the discrepancy ?
Shold I then use the first continuum assuming that the synthesis gives me the "right " F1 values? -------Mensagem original-------
Hi
Re your synthesis question: how are you measuring the F1 of the
synthesised vowel? It may be that it is difficult to measure accurately
(either manually or by some automated approach such as LPC) in the F1
region due to the fact that F1 does not necessarily correspond to a
spectral peak (rather, spectral peaks in that region will be harmonics
of F0). To give a fuller answer, I'd need to know details of the
synthesis/analysis approach.
Martin | |||
|
|
