Good morning, I'm Filippo Battaglia from Visilab Research
Center - University of Messina - Italy.
We're working from several days to port PocketSphinx under
Nanodesktop. We've investigating about the troubles that we
have found previously (no words recognized in pocketsphinx_tidigits).
We've done deep analysis, comparing the behaviour of
Nanodesktop version and of the Linux version of Sphinx.
We've found this trouble.
In the module ngram_search.c, routine ngram_search_find_exit,
under Nanodesktop the array
ngs->bp_table_idx [frame_idx]
contains only ZERO values. Under Linux, there are several not-zero values...
We believe that this anomaly is connected with another fact.
We have debugged the routine ngram_search_step in the same
module. And we've found that a printf of the value
ngs->bpidx
returns always a zero value under Nanodesktop (not zero value
under Linux).
So, the problem should be before this routine.
We cannot find who invokes this routine (I believe that there
is some kind of pointer to function that invokes indirectly
ngram_search_step, but I don't understand what is exactly
the calling routine).
Is it possible to talk with one of the developers of the
library ? Has anyone his mail address ?
Thanks in advance for your help..
Filippo Battaglia
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Good morning, I'm Filippo Battaglia from Visilab Research
Center - University of Messina - Italy.
We're working from several days to port PocketSphinx under
Nanodesktop. We've investigating about the troubles that we
have found previously (no words recognized in pocketsphinx_tidigits).
We've done deep analysis, comparing the behaviour of
Nanodesktop version and of the Linux version of Sphinx.
We've found this trouble.
In the module ngram_search.c, routine ngram_search_find_exit,
under Nanodesktop the array
ngs->bp_table_idx [frame_idx]
contains only ZERO values. Under Linux, there are several not-zero values...
We believe that this anomaly is connected with another fact.
We have debugged the routine ngram_search_step in the same
module. And we've found that a printf of the value
ngs->bpidx
returns always a zero value under Nanodesktop (not zero value
under Linux).
So, the problem should be before this routine.
We cannot find who invokes this routine (I believe that there
is some kind of pointer to function that invokes indirectly
ngram_search_step, but I don't understand what is exactly
the calling routine).
Is it possible to talk with one of the developers of the
library ? Has anyone his mail address ?
Thanks in advance for your help..
Filippo Battaglia
Also if bpidx is always zero, it just means that no words are being recognized ... there is some deeper error which must be resulting in this.
Yes, please send me e-mail about this, I find these message boards to be difficult at times. The address is
dhuggins ++AT++ cs ++DOT++ cmu ++DOT++ edu
You can also post questions to the cmusphinx-sdmeet mailing list.
I have sent a mail to your address.
I hope that you receive it.
Thank you for your help
Filippo