Hi,
I am working on a MELPe vocoder at a rate higher than 2400bps. I just was trying
to increase the number of LPC's to 12. And the code some how introduces a
tonal noise into the speech. If I increase the number of LPC's further, the
noise in the reconstructed speech further increases. I give it a 21sec speech
and there is a tone at about 1.65s and one at 14.5s. Did anybody try to increase
the number of LPC's and come across something like this. Can you suggest me
as to where the bug could be?
And also can anybody give me or point me to a source code that converts
LPC's (prediction coeffs) to LSF's for any order.
Thank you,
Aditi.