I am on Ubuntu 11.10 and following the tutorial for training data using
sphinx3, sphinxbase, sphinxtrain and the an4 data sample. Everything worked
fine until the step transforming the signal into a sequence of feature vectors
by typing the following command :
ERROR: "wave2feat.c", line 655: Cannot read /home/dev/Documents/Developpement/Audio/CMUSphinx/tutorial/an4/wav/an4_clstk/fash/an251-fash-b.sph
FATAL_ERROR: "wave2feat.c", line 90: error converting files...exiting
Indeed, in /an4/wav/an4_clstk/fash there is no .sph files. I guess the script
is supposed to generate them but it didn't. Could you please tell me more
about this step ? What do you think is missing ?
Thanks advance
Cédric K
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
SPH files contain audio. They are not generated, they must be downloaded from
the website. Please check that you downloaded the archive and upacked it
properly.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thanks for your reply.
I have checked the tarball package and there are no sph files. However, in
sphinx_train.cfg file, I changed the $CFG_WAVFILE_EXTENSION and
$CFG_WAVFILE_TYPE as 'wav' and 'mswav' ( was sph and nist ) and I could run
the command properly.
However, during the execution, I get a memory error similar to this one : htt
ps://sourceforge.net/tracker/?func=detail&aid=3450890&group_id=1904&atid=10190
4.
buffer overflow detected
In the above link, you propose to use sphinx_ve instead of wave2feat to build
the features. How can I switch wave2feat to sphinx_ve ?
Thanks in advance
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello everyone
I am on Ubuntu 11.10 and following the tutorial for training data using
sphinx3, sphinxbase, sphinxtrain and the an4 data sample. Everything worked
fine until the step transforming the signal into a sequence of feature vectors
by typing the following command :
I get the following error :
Indeed, in /an4/wav/an4_clstk/fash there is no .sph files. I guess the script
is supposed to generate them but it didn't. Could you please tell me more
about this step ? What do you think is missing ?
Thanks advance
Cédric K
SPH files contain audio. They are not generated, they must be downloaded from
the website. Please check that you downloaded the archive and upacked it
properly.
Thanks for your reply.
I have checked the tarball package and there are no sph files. However, in
sphinx_train.cfg file, I changed the $CFG_WAVFILE_EXTENSION and
$CFG_WAVFILE_TYPE as 'wav' and 'mswav' ( was sph and nist ) and I could run
the command properly.
However, during the execution, I get a memory error similar to this one : htt
ps://sourceforge.net/tracker/?func=detail&aid=3450890&group_id=1904&atid=10190
4.
In the above link, you propose to use sphinx_ve instead of wave2feat to build
the features. How can I switch wave2feat to sphinx_ve ?
Thanks in advance
You can find all the information on training in the tutorial
http://cmusphinx.sourceforge.net/wiki/tutorialam
It also contains a links to tarball with sph files. Please make sure you read
it before you start training.
Alright, I guess I was reading a previous version of the tutorial ( http://ww
w.speech.cs.cmu.edu/sphinx/tutorial.html ).
I am gonna start with the tutorial you linked.
Thanks and sorry about that.
Cédric