Hi Simon,
I did the upsampling because I was using the original HTS recipe and did not want to change the parameters about frequency and window size before I got a working voice (to make sure I was not changing too many things at once in the scripts to track where the quality change might come from)
I plan to update those to use the original frequency when I get a decently working voice
do you have suggestions on proper parameters to use for the 16kHz case, and do you think that upsampling would result in a lower-quality voice than using the samples as they are?
On 28 July 2017 at 10:44:39, Simon King (simon.king@xxxxxxxx) wrote:
Maël Primet wrote:
>
> I’m starting with 16kHz samples that I upsampled to 48kHz
>
You understand that this upsampling will result in a signal that still
has no energy above the Nyquist frequency of the original signal,
right? (If not, just take look at a spectrogram of the upsampled
signal) That will have consequences when you parameterise the upsampled
version.
Simon
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.