just guessing : shouldn't it be -d ("end condition") ?
Yufeng Hao a écrit :
And now, "-e 0.01" option is applied for try. now more
warnings appears temporayly.
Heiga, do you have some papre about mgc-lsp feature extraction? :)
YF HAO
On Nov 19, 2007 6:11 PM, Yufeng Hao <
yufenghao@xxxxxxxxx>
wrote:
after
I changed the coefficients as you said, shown in below:
$(MGCEP) -a $(FREQWARP) -g $(GAMMA) -m
$(MGCORDER) -f 0.01 -l $(FFTLEN) -o 4 | \
New more errors appeared:
Extracting MGC-LSP coefficients
from raw/cmu_us_arctic_slt_a0001.raw
mgcep : Invalid option 'f'!
mgcep - mel-generalized cepstral analysis
usage:
mgcep [ options ] [ infile ] > stdout
options:
-a a : alpha [0.35]
-g g : gamma [0]
-m m : order of mel-generalized cepstrum [25]
-l l : frame length [256]
-o o : output format [0]
0 (c~0...c~m)
1 (b0...bm)
2 (K~,c~'1...c~'m)
3 (K,b'1...b'm)
4 (K~,g*c~'1...g*c~'m)
5 (K,g*b'1...g*b'm)
(level 2)
-i i : minimum iteration [2]
-j j : maximum iteration [30]
-d d : end condition [
0.001]
-p p : order of recursions [l-1]
-e e : small value added to periodgram [0]
-h : print this message
infile:
windowed sequence (float) [stdin]
stdout:
mel-generalized cepstrum (float)
notice:
if g >= 1.0, g = -1 / g
SPTK: version 3.1
CVS Info: $Id: mgcep.c,v 1.19 2007/09/30 16:20:31 heigazen Exp $
Extracting MGC-LSP coefficients from raw/cmu_us_arctic_slt_a0002.raw
mgcep : Invalid option 'f'!
There is no "-f" options in mgcep in SPTK 3.1. :(
And next?
YFHAO
|