Hi all, I am working on to do some simple acoustic model adaptation to improve
Speech recognition according to the link http://cmusphinx.sourceforge.net/wik
i/tutorialadapt in that
i done the following steps successfully.
_ 1.Creating an adaptation corpus
2.Recording your adaptation data
3.Generating acoustic feature files
4.Converting the sendump and mdef files
_
I found all the files need for the process in my Directory.
However i found an issue with Accumulating observation counts step.
my console shows an error
stat_retry(arctic_0001..mfc) failed
ERROR: "corpus.c", line 1555: MFCC read of arctic_0001..mfc failed. Retrying
after sleep...
FATAL_ERROR: "corpus.c", line 1560: Failed to get the files after 100 retries
(about 300 seconds)
may i know which is missing... what mistake i done..
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I saw the check mark option "Execute" as unchecked in the arctic_0001.mfc file
permission window which will give permissions to allow a file to be run as a
program .is it the reason?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
But while collecting Statistics using adaption data, I get following error. (
note the file extension with two periods instead of one period )
stat_retry(arctic_0001..mfc) failed ERROR: "corpus.c", line 1555: MFCC read
of arctic_0001..mfc failed. Retrying after sleep... FATAL_ERROR: "corpus.c",
line 1560: Failed to get the files after 100 retries (about 300 seconds)
Command used for collecting statistics is as below.
Earlier i was thinking this issue was due to insufficient permissions to read
the file. But that is not it. now these files have full permissions ( read,
write and execute). Still same error though.
I am unable to figure what part of the code is making the filenames to be
appear as arctic_0001..mfc
Any help on figuring out this issue is appreciated.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
MFC file name is constructed from the line in fileids file and .mfc extension.
YOu have extra symbols in fileids file which you need to cleanup (most likely
windows-style newlines).
You are also recommended to use latest sphinxtrain version. It will solve many
of your problems automatically.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
nshmyrev,
You recommended using latest snapshot of sphinx train. Is there a win 32
binary of the latest snapshot available for download some where? I am from a
pure java background, if this was java open source , i would have gotten the
src and compiled it, but i am kind of lost with the C/C++ stuff.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello everyone,
I have the same problem with achuamr. I have changed the corpus.c and
installed the latest version of sphinxtrain, but it does not work either..
Who can help me? Thanks a lot!!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Nancyxll, do not edit any source code. Latest version of sphinxtrain should
just work. Are you sure you installed it properly?
Which exact place did you take it?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It's not that simple. In latest versions cepext is mfc by default. If you are
using old one you possible will have more problems not just this one. So it's
highly recommended to upgrade.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It said they are the latest available releases. But it still has that problem.
I have solved the problem now. It can work if you change the function. The
reason why I didn't succeed because I forgot to replace my older bw etc. bin
files in my working directory.
So, if the others will meet the same problem, just change the function as
above. And DON'T FORGET TO REPLACE your old bin files.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi ALL,
I met a problem when I do "Creating transformation with MLLR".
First, there is no instruction above talked about MLLR, I don't knwo MLLR is
about what...
Second, I haven't found sphinx3 under /usr/local/share/ (but I have
pockentsphinx under share folder, however, there is no
hub4_cd_continuous_8gau_1s_c_d_dd under hmm folder).
What should I do?
Since I know little about voice recognition, this is my first time to use
Sphinx. So I met lots of problems... Could you do me a favor to solve this
transformation problem? Many thanks!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
--------I have solved the problem now. It can work if you change the
function. The reason why I didn't succeed because I forgot to replace my older
bw etc. bin files in my working directory--------
that's why i told to do the steps again from Adapting the acoustic model
from the site.
any way the issue is solved na.. happy
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
First, there is no instruction above talked about MLLR, I don't knwo MLLR is
about what... Second, I haven't found sphinx3 under /usr/local/share/ (but I
have pockentsphinx under share folder, however, there is no
hub4_cd_continuous_8gau_1s_c_d_dd under hmm folder). What should I do?
Hi Shahlashahlu,
Should I use the mllr_solve command just under my working directory?
I have a hub4wsj_sc_8k folder under my working directory.
I run this command twice.
First, I run it just under my working directory (it contains my .wav and .mfc
files), it succeeded.
Second, I run it under the hub4wsj_sc_8k folder in my working directory, it
failed.
So I think, we should run it under a directory that has .wav and .mfc files,
right?
Another thing is:
when I run map_adapt under my current working directory, it said: "map_adapt:
command not found".
I don't know why...
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
First, I run it just under my working directory (it contains my .wav and
.mfc files), it succeeded.
i think that's fine.
Second, I run it under the hub4wsj_sc_8k folder in my working directory, it
failed.
because the path hub4wsj_sc_8k/means representing from your working
directory which contains the .wav and other files.So it will not work. First
one is enough.
when I run map_adapt under my current working directory, it said:
"map_adapt: command not found".
Thank you shahlashahlu!
I am now having a question about how to run it...
Actually, I wan to create a model for cell phone using. But right now, I just
do all the things on my PC. So I think I should use Sphinx4 at present.
It seems I have adjusted my model.
In the tutorial, it said 'Now test your accuracy to see it's good. ' But I
don't know how to test..
I have installed sphinxbase, pocketshpinx, and build sphinx4 (for sphinx4, I
don't know whether I should do further, but I built it already) on my Ubuntu,
which runs in VMware.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi All,
I have already adjusted my acoustic model, and now I want to use it.
My thought is:
1. using the Sphinx 'hellngram' demo and change some parameters and paths so that it can conclude my acoustic model, including dictionary.
2. should I change "hellongram.config.xml" file?
In this file, I found
I am facing the same problem you faced earlier.
"stat_retry(arctic_0001..mfc) failed ERROR: "corpus.c", line 1555: MFCC read
of arctic_0001..mfc failed. Retrying after sleep... FATAL_ERROR: "corpus.c",
line 1560: Failed to get the files after 100 retries (about 300 seconds)"
I have gone through the post but i didn't understand what you really mean by
statement:
"So, if the others will meet the same problem, just change the function as
above. And DON'T FORGET TO REPLACE your old bin files. "
my query for you and OTHERS are:
1.which 'fn' you guyz are suggesting.
2."bw etc. bin" which files you are suggestings.
I already pasted bw,map_adapt and mk_s2sendump files into my working
directory..what other files you are referring.
Please elaborate and help.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
Hi all, I am working on to do some simple acoustic model adaptation to improve
Speech recognition according to the link http://cmusphinx.sourceforge.net/wik
i/tutorialadapt in that
i done the following steps successfully.
_ 1.Creating an adaptation corpus
2.Recording your adaptation data
3.Generating acoustic feature files
4.Converting the sendump and mdef files
_
I found all the files need for the process in my Directory.
However i found an issue with Accumulating observation counts step.
my console shows an error
FATAL_ERROR: "corpus.c", line 1560: Failed to get the files after 100 retries
(about 300 seconds)
may i know which is missing... what mistake i done..
the error was occurred when i run the command
the error line shows
ERROR: "corpus.c", line 1555: MFCC read of arctic_0001..mfc **failed. Retrying after sleep... **
the error line shows the file name as ..mfc extension. actually the file
is **.mfc ** extension.
may i know which is missing... what mistake i done..
my feat.params is
I saw the check mark option "Execute" as unchecked in the arctic_0001.mfc file
permission window which will give permissions to allow a file to be run as a
program .is it the reason?
Think more how additional symbol can appear in mfc file name.
MFC files were generated with .mfc extension as below
arctic_0001.mfc
arctic_0002.mfc
...( upto )
arctic_0020.mfc
These MFC files were created using following command.
But while collecting Statistics using adaption data, I get following error. (
note the file extension with two periods instead of one period )
Command used for collecting statistics is as below.
Earlier i was thinking this issue was due to insufficient permissions to read
the file. But that is not it. now these files have full permissions ( read,
write and execute). Still same error though.
I am unable to figure what part of the code is making the filenames to be
appear as arctic_0001..mfc
Any help on figuring out this issue is appreciated.
MFC file name is constructed from the line in fileids file and .mfc extension.
YOu have extra symbols in fileids file which you need to cleanup (most likely
windows-style newlines).
You are also recommended to use latest sphinxtrain version. It will solve many
of your problems automatically.
nshmyrev,
You recommended using latest snapshot of sphinx train. Is there a win 32
binary of the latest snapshot available for download some where? I am from a
pure java background, if this was java open source , i would have gotten the
src and compiled it, but i am kind of lost with the C/C++ stuff.
Hello everyone,
I have the same problem with achuamr. I have changed the corpus.c and
installed the latest version of sphinxtrain, but it does not work either..
Who can help me? Thanks a lot!!
Nancyxll, do not edit any source code. Latest version of sphinxtrain should
just work. Are you sure you installed it properly?
Which exact place did you take it?
simply add
for bw to resolve this problem
Yaksea
It's not that simple. In latest versions cepext is mfc by default. If you are
using old one you possible will have more problems not just this one. So it's
highly recommended to upgrade.
Thank you guys! I am so appreciated your quickly reply.
I downloaded all of the softwares from the website:
http://cmusphinx.sourceforge.net/wiki/download/
It said they are the latest available releases. But it still has that problem.
I have solved the problem now. It can work if you change the function. The
reason why I didn't succeed because I forgot to replace my older bw etc. bin
files in my working directory.
So, if the others will meet the same problem, just change the function as
above. And DON'T FORGET TO REPLACE your old bin files.
Hi ALL,
I met a problem when I do "Creating transformation with MLLR".
First, there is no instruction above talked about MLLR, I don't knwo MLLR is
about what...
Second, I haven't found sphinx3 under /usr/local/share/ (but I have
pockentsphinx under share folder, however, there is no
hub4_cd_continuous_8gau_1s_c_d_dd under hmm folder).
What should I do?
Since I know little about voice recognition, this is my first time to use
Sphinx. So I met lots of problems... Could you do me a favor to solve this
transformation problem? Many thanks!
@Xianglian Zhang:
that's why i told to do the steps again from Adapting the acoustic model
from the site.
any way the issue is solved na.. happy
@Xianglian Zhang:
you are worrying about the path issue?
if you are in current directory , try the below
Example:
lets try..
Thank you so.....o much Shahlashahlu, I will try it at once. Thank you!
Hi Shahlashahlu,
Should I use the mllr_solve command just under my working directory?
I have a hub4wsj_sc_8k folder under my working directory.
I run this command twice.
First, I run it just under my working directory (it contains my .wav and .mfc
files), it succeeded.
Second, I run it under the hub4wsj_sc_8k folder in my working directory, it
failed.
So I think, we should run it under a directory that has .wav and .mfc files,
right?
Another thing is:
when I run map_adapt under my current working directory, it said: "map_adapt:
command not found".
I don't know why...
@Xianglian Zhang:
i think that's fine.
because the path hub4wsj_sc_8k/means representing from your working
directory which contains the .wav and other files.So it will not work. First
one is enough.
please add a ./ before the map_adapt.
eg:
./map_adapt \
-meanfn hub4wsj_sc_8k/means \
-varfn hub4wsj_sc_8k/variances \
-mixwfn hub4wsj_sc_8k/mixture_weights \
-tmatfn hub4wsj_sc_8k/transition_matrices \
-accumdir . \
-mapmeanfn hub4wsj_sc_8kadapt/means \
-mapvarfn hub4wsj_sc_8kadapt/variances \
-mapmixwfn hub4wsj_sc_8kadapt/mixture_weights \
-maptmatfn hub4wsj_sc_8kadapt/transition_matrices
it will work , if you are in working directry(which contains the .wav and mfc
files).
let see..
Thanks.
Yes it works, shahlashahlu!
Thank you !!!:)))
@Xianglian Zhang: ok.. may i know if any other problem occurs !!!:))) . Happy
coding.. May i know which plat form you are working.
Are you creating the acoustic model for Pocket-Sphinx?
Thank you shahlashahlu!
I am now having a question about how to run it...
Actually, I wan to create a model for cell phone using. But right now, I just
do all the things on my PC. So I think I should use Sphinx4 at present.
It seems I have adjusted my model.
In the tutorial, it said 'Now test your accuracy to see it's good. ' But I
don't know how to test..
I have installed sphinxbase, pocketshpinx, and build sphinx4 (for sphinx4, I
don't know whether I should do further, but I built it already) on my Ubuntu,
which runs in VMware.
Does that mean I should write a program by myself, which includes the adjusted
acoustic model?
Hi All,
I have already adjusted my acoustic model, and now I want to use it.
My thought is:
1. using the Sphinx 'hellngram' demo and change some parameters and paths so that it can conclude my acoustic model, including dictionary.
2. should I change "hellongram.config.xml" file?
In this file, I found
<component name="lexTreeLinguist" type="edu.cmu.sphinx.linguist.lextree.LexTreeLinguist">
<property name="logMath" value="logMath">
<property name="acousticModel" value="wsj">
<property name="languageModel" value="trigramModel">
<property name="dictionary" value="dictionary">
<property name="addFillerWords" value="false">
<property name="fillerInsertionProbability" value="1E-10">
<property name="generateUnitStates" value="false">
<property name="wantUnigramSmear" value="true">
<property name="unigramSmearWeight" value="1">
<property name="wordInsertionProbability" value="${wordInsertionProbability}">
<property name="silenceInsertionProbability" value="${silenceInsertionProbability}">
<property name="languageWeight" value="${languageWeight}">
<property name="unitManager" value="unitManager">
</property></property></property></property></property></property></property></property></property></property></property></property></property></component>
talked about acoustic model and dictionary.
But I don't know how to point it to my own acoustic model.
Could you do me a favor?
Many thanks!!
hii nancyxll
I am facing the same problem you faced earlier.
"stat_retry(arctic_0001..mfc) failed ERROR: "corpus.c", line 1555: MFCC read
of arctic_0001..mfc failed. Retrying after sleep... FATAL_ERROR: "corpus.c",
line 1560: Failed to get the files after 100 retries (about 300 seconds)"
I have gone through the post but i didn't understand what you really mean by
statement:
"So, if the others will meet the same problem, just change the function as
above. And DON'T FORGET TO REPLACE your old bin files. "
my query for you and OTHERS are:
1.which 'fn' you guyz are suggesting.
2."bw etc. bin" which files you are suggestings.
I already pasted bw,map_adapt and mk_s2sendump files into my working
directory..what other files you are referring.
Please elaborate and help.