I am facing a very odd problem, and that is the training suite takes too much time in creating the melcep feature files. I have a single wave file (conforming to the standards) having an utterance 'Precarious', and the .dict file also exists (i created it).
Whats the reason for this time delay (an hour and not done yet).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Are you sure your file is reallly existed(or if the path is correctly specified)? Some of the tools in SphinxTrain will retry to get the mfcc files if it doesn't exit.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
what are mfcc files...the system hangs up when i execute the script 'make_feat'. The script in turn calls wave2feat and then the system is literally blocked, last night i tried using the script for an utterance 'ONE' and doing all the pre-requisites as mentioned in the documentation but still after 90min the process was not complete, there was 2.1
% mem usage and 99.1% CPU usage.
I am still at a loss about the problem and am trying to understand the code in the wave2feat library. Is there an easier solution or some checklist of files which i can get.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
What are the mfcc files?
MFCC are mel frequency cepstral coefficient, one of the standard type of encoding of wave form. wave2feat is the program that turn the waveform to MFCC.
From your mail, it seems to me that wave2feat was trying to find a wave file but failed. Now SphinxTrain will retry to get the file again after 5 minutres. This is very likely to be reason of why you can't get it work.
Go to take a look of the log generated by makefeat and see whether there are anything such as "Retry..."
in your log. If there is, check the path of your wave file and check whether you have specified them correctly. You need to specifiy -cepdir and -cepext correctly in your script.
A side-note of script and code tracing. Try to understand the script and the log first. Directly tracing tools like will cause you very long time (Well, it is fun though. :-)
Arthur
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Well i got the problem sorted out, it was that the wav input file was not in the correct format (nist) so now i am giving it raw files.
Now another problem has crept up and that is that the tree building does not function properly, no trees are made by the script and the error which is generated in the log is:
FATAL ERROR: "ckd_alloc.c" , line 85: Calloc failed from main.c (699).
Another query how are we supposed to specify the triphones (there is a utility dict2tri) but how should i tell the scripts about the triphone file and how to create one.
regards
BILAL AHMED
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am facing a very odd problem, and that is the training suite takes too much time in creating the melcep feature files. I have a single wave file (conforming to the standards) having an utterance 'Precarious', and the .dict file also exists (i created it).
Whats the reason for this time delay (an hour and not done yet).
Are you sure your file is reallly existed(or if the path is correctly specified)? Some of the tools in SphinxTrain will retry to get the mfcc files if it doesn't exit.
what are mfcc files...the system hangs up when i execute the script 'make_feat'. The script in turn calls wave2feat and then the system is literally blocked, last night i tried using the script for an utterance 'ONE' and doing all the pre-requisites as mentioned in the documentation but still after 90min the process was not complete, there was 2.1
% mem usage and 99.1% CPU usage.
I am still at a loss about the problem and am trying to understand the code in the wave2feat library. Is there an easier solution or some checklist of files which i can get.
What are the mfcc files?
MFCC are mel frequency cepstral coefficient, one of the standard type of encoding of wave form. wave2feat is the program that turn the waveform to MFCC.
From your mail, it seems to me that wave2feat was trying to find a wave file but failed. Now SphinxTrain will retry to get the file again after 5 minutres. This is very likely to be reason of why you can't get it work.
Go to take a look of the log generated by makefeat and see whether there are anything such as "Retry..."
in your log. If there is, check the path of your wave file and check whether you have specified them correctly. You need to specifiy -cepdir and -cepext correctly in your script.
A side-note of script and code tracing. Try to understand the script and the log first. Directly tracing tools like will cause you very long time (Well, it is fun though. :-)
Arthur
Well i got the problem sorted out, it was that the wav input file was not in the correct format (nist) so now i am giving it raw files.
Now another problem has crept up and that is that the tree building does not function properly, no trees are made by the script and the error which is generated in the log is:
FATAL ERROR: "ckd_alloc.c" , line 85: Calloc failed from main.c (699).
Another query how are we supposed to specify the triphones (there is a utility dict2tri) but how should i tell the scripts about the triphone file and how to create one.
regards
BILAL AHMED