htk 3.5

[kothemel]

Καλημέρα παιδιά!

 

Έχει ασχοληθεί κανείς με τη νέα έκδοση του ΗΤΚ 3.5 πάνω σε DNNs;

Έχω κολλήσει στο στάδιο του pre-training. Παραθέτω παρακάτω αναλυτικά το μήνυμα λάθους που μου πετάει.

 

Στο φορουμ του voxforge:

http://www.voxforge.org/home/forums/message-boards/speech-recognition-engines/

 

I have stuck in the pre-training step of DNN-HMM.

First of all, the whole database (.wav, .mfc, .mlf files) is stored on an ext. drive. According to the HTKBook page 68 in order to proceed with descriminative pre-training you must execute:

HNTrainSGD -C config.basic -C config.pretrain -H dnn3/init/models -M dnn3  -S dnn.train.scp -N dnn.hv.scp -l LABEL -I train.mono.aligned.mlf  hmm_mono/monolist

After creating the config.basic, config.pretrain, dnn.train.scp, dnn.hv.scp and running the command I get the following error:

 

Epoch 1 ******************************

Processing training set...

 ERROR [+8925]  LoadOneUtt: Phone label in the label file does not match the state level definition

FATAL ERROR - Terminating program ../../bin.cpu/HNTrainSGD

 

Well i guess the problem is occurred by the massive label file.

The train.mono.aligned.mlf, as I defined it, contains the paths to other mlfs. I quote a small part of it:

#!MLF!#
"/media/kothemel/Elements1/training_set/S01/ISOLATED/mic1/s01mic1.mlf"
"/media/kothemel/Elements1/training_set/S01/ISOLATED/mic2/s01mic2.mlf"

and goes on..

On the ext. drive the s01mic1.mlf contains the acoustic events as labels with the utterance that is taking place. I also quote a part of it:

#!MLF!#
"mic1/S01-eventid1-mic1.lab"
si
.
"mic1/S01-eventid2-mic1.lab"
kn
.
"mic1/S01-eventid3-mic1.lab"
si
.
"mic1/S01-eventid4-mic1.lab"
ds
.
"mic1/S01-eventid5-mic1.lab"
si
.
"mic1/S01-eventid6-mic1.lab"
ds
.