VoxForge
VoxForge
Hi,
I'm trying to perform silence removal on utterances using htk but I keep getting errors.I tried using HCopy with the silence parameters in the config file.
Can someone help me with this?
~rajesh
--- (Edited on 8/12/2009 12:32 am [GMT-0500] by ) ---
> Can someone help me with this?
Sure, with the answer that has the same amount information as the question.
--- (Edited on 8/12/2009 2:17 am [GMT-0500] by nsh) ---
ok..so the parameters I used are as follows
SOURCEKIND = WAVEFORM
TARGETKIND = WAVEFORM
USESILDET = T
MEASURESIL = T
and I said
HCopy -C config src.wav targt.wav
am I missing something?
--- (Edited on 8/12/2009 3:57 am [GMT-0500] by ) ---
HTK only applies the silence detector if you extract features (do a framing processing). Silence detector is disabled when you copy from wav to wav, this way you can only chop the files.
--- (Edited on 8/13/2009 6:30 pm [GMT-0500] by nsh) ---
So from what you are saying if I am going from SOURCEFORMAT=NIST to TARGETKIND=MFCC_0 and have USESILDET=T silence detection/removal should occur?
My configuration is as follows:
SOURCEFORMAT = NIST
TARGETKIND = MFCC_0
TARGETRATE = 100000.0
WINDOWSIZE = 250000.0
USEHAMMING = T
PREEMCOEF = 0.97
NUMCHANS = 26
CEPLIFTER = 22
NUMCEPS = 12
ENORMALISE = F
USESILDET = T
MEASURESIL = T
OUTSILWARN = T
However, when I run HCopy I get the same output as I do when I run HCopy with the same configuration file minus the last three lines (USESILDET, MEASURESIL, OUTSILWARN).
Thoughts?
--- (Edited on 9/23/2009 8:04 pm [GMT-0500] by Visitor) ---