Hello everyone,

I am currently in the process of making an HTK model using the TIMIT database, to learn how HTK works in depth.

So I am currently at the step 3.2.3 of the HTK Book.

My command line is :

"HVite -l '*' -o SWT -b silence -C hcompv.conf  -a -H hmm14/macros -H hmm14/hmmdefs -i aligned.mlf -m -t 250.0 -y lab -I wrd.mlf -S train.scp output.dict monophones"

So, in my wrd.mlf, I have 9240 pseudo-label-file entries, like for instance :

"*/timit_train_DR3_MTPG0_SX393.lab"
1787500 3446250 she

3446250 6775000 uses

6775000 9558750 both

10111875 14443750 names

15021250 22425000 interchangeably

.

but in the output of HVite, aligned.mlf, I only have 9126 entries !

(The corresponding one, for instance :

"*/timit_train_DR3_MTPG0_SX393.lab"

sil

sh

iy

sp

y

uw

z

ih

z

sp

vcl

b

ow

th

sp

n

ey

m

z

sp

ih

n

vcl

t

er

ch

ey

n

jh

ax

vcl

b

l

iy

sp

sil

.

)

So of course, later when I have to retrain using HERest, it doesn't work complaingin that a label file is not found.

Is this a bug ? Have I missed something important ?

Thanks everybody.

--- (Edited on 6/26/2013 2:55 am [GMT-0500] by ) ---