Hi all,
I'm GSoC student, which works on project Simon SPHINX support http://www
.google-melange.com/gsoc/proposal/review/google/gsoc2012/stvlad/1 . And as a
part of my project i need to convert data from internal Simon format to format
acceptable to SPHINX and then train own acoustic model from this data.
And now, I had the following problem:
when I try to recognize something, using the created model, vocabulary and
jsgf grammar, I'm on almost any query about getting the following response:
"READY ....
Listening ...
Stopped listening, please wait ...
INFO: cmn_prior.c (121): cmn_prior_update: from <12.52 -1.49 -0.64 0.21 -0.56
0.18 -0.19 -0.04 -0.33 -0.07 -0.18 -0.24 -0.12>
INFO: cmn_prior.c (139): cmn_prior_update: to <12.66 -1.44 -0.63 0.20 -0.56
0.16 -0.19 -0.05 -0.31 -0.06 -0.15 -0.23 -0.13>
INFO: fsg_search.c (1032): 69 frames, 543 HMMs (7/fr), 1361 senones (19/fr),
125 history entries (1/fr)
ERROR: "fsg_search.c", line 1104: Final result does not match the grammar in
frame 69
000000086: (null)
"
can you tell me please: what am I doing wrong?
example of generated model: http://db.tt/SCVmSpw1
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The message says that the text you are trying to recognize does not match the
grammar. The JSGF grammar you wrote accepts only several sentences from a test
set:
Means that the phrase MUST start only from a words from a first subset and
then MUST go one of the words from second subset. To write a proper grammar
you need to use repetition symbols like * or + to tell the recognizer that
words can repeat in any order if you meant that originally.
You can learn more details from JSGF grammar documentation
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2012-07-19
Means that the phrase MUST start only from a words from a first subset and
then MUST go one of the words from second subset.
Yes that's exactly what I want to reach at the stage of recognition.
The JSGF grammar you wrote accepts only several sentences from a test set:
so testing set should contain only those records that satisfy the grammar? And
this is the reason why I get such a response to almost all queries (in the
correct format)?
Also, I have a question that I have not seen in the JSGF documentation:
Can one file contain multiple grammars?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all,
I'm GSoC student, which works on project Simon SPHINX support http://www
.google-melange.com/gsoc/proposal/review/google/gsoc2012/stvlad/1 . And as a
part of my project i need to convert data from internal Simon format to format
acceptable to SPHINX and then train own acoustic model from this data.
And now, I had the following problem:
when I try to recognize something, using the created model, vocabulary and
jsgf grammar, I'm on almost any query about getting the following response:
"READY ....
Listening ...
Stopped listening, please wait ...
INFO: cmn_prior.c (121): cmn_prior_update: from <12.52 -1.49 -0.64 0.21 -0.56
0.18 -0.19 -0.04 -0.33 -0.07 -0.18 -0.24 -0.12>
INFO: cmn_prior.c (139): cmn_prior_update: to <12.66 -1.44 -0.63 0.20 -0.56
0.16 -0.19 -0.05 -0.31 -0.06 -0.15 -0.23 -0.13>
INFO: fsg_search.c (1032): 69 frames, 543 HMMs (7/fr), 1361 senones (19/fr),
125 history entries (1/fr)
ERROR: "fsg_search.c", line 1104: Final result does not match the grammar in
frame 69
000000086: (null)
"
can you tell me please: what am I doing wrong?
example of generated model:
http://db.tt/SCVmSpw1
The message says that the text you are trying to recognize does not match the
grammar. The JSGF grammar you wrote accepts only several sentences from a test
set:
Means that the phrase MUST start only from a words from a first subset and
then MUST go one of the words from second subset. To write a proper grammar
you need to use repetition symbols like * or + to tell the recognizer that
words can repeat in any order if you meant that originally.
You can learn more details from JSGF grammar documentation
http://java.sun.com/products/java-
media/speech/forDevelopers/JSGF/
Also, I have a question that I have not seen in the JSGF documentation:
Can one file contain multiple grammars?
No
A single file defines a single grammar.