I am trying to run MLLR adaptation in sphinx. When I try with "-cd_only no" option, I get error messeges "Error: final state not reached". Sometimes I also get negative convergence ratio in the adaptation process. When I do the decoding with these adapted models, the recognition is really bad, sometimes the whole sentences are gone (compelete sentence deleted).
When I try with "-cd_only yes" option, I get segmentation fault during mllr_transform.
Can somebody please help me with this problem? I tried with different set of adaptation data but the same happens. I am using TIMIT dbase.
many thanks,
abhishek.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi ,
I am trying to run MLLR adaptation in sphinx. When I try with "-cd_only no" option, I get error messeges "Error: final state not reached". Sometimes I also get negative convergence ratio in the adaptation process. When I do the decoding with these adapted models, the recognition is really bad, sometimes the whole sentences are gone (compelete sentence deleted).
When I try with "-cd_only yes" option, I get segmentation fault during mllr_transform.
Can somebody please help me with this problem? I tried with different set of adaptation data but the same happens. I am using TIMIT dbase.
many thanks,
abhishek.
Can you provide a log and a stacktrace of a crash?
http://fedoraproject.org/wiki/StackTraces