I want to do some tests on the plateform which is like wavfile demo. my demo name is hopewavdemo."ant" is successful.But,when execute hopewavfile.jar ,the result is "blank" as follows:
Hi Will:
Thank you.However,because my poor English,I do not express that clearly.Now,I will say more.
I have done one demo named "hope"which is liked "HelloDigits"demo.The acoustic model was also trained by myself.I do not use Sphinx4's demos and models.
The "hope" demo worked well.But It demands that the "input files" should be from microphone.So, I want the input file is wavfile which is like "wavfile"demo.
I do one demo named "hopewavfile",but, the results is not what I want to get.I do not konw where the problem have occured.
Hope you could understand what I want to express. Thanks a lot in advance.
Maggie
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
There could be any number of problems here. I'm going to guess that your lH.wav file does not contain someone speaking just digits, but that it contains other words, and that your config file and grammar are still set up for digits.
However, the more important thing here is to understand what task/problem you are trying to support. If you describe this, we might be able to point you down another path.
Will
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Will:
Thanks a lot for your advice.I will say the task in more detail.
Shortly,I want to do one demo like wavfile.This demo's name is hopewavfile,which contains similar files with wavfile.Such as config.xml ,hopewavfile,hopewavfile.Manifest,hopewavfile.gram，IH.wav
hopewavfile.gram is:
JSGF V1.0;
/*
* JSGF Digits Grammar for Hello World example /
grammar hope;
public <voice> = (EH | IH | UH | YU ) * ;
The other files are also modified.
The results is BLANK...
I do not konw how to solve it.
Maggie You
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
1) You are using the digits models (i.e., things really good at recognizing only numbers) with a grammar containing no digits.
2) Even if we get the config file set up to use more general models, at least one of the words in the grammar, IH, is not in the pronunciation lexicon.
An alternative approach for you might be to start with the demo/sphinx/helloworld demo, using its helloworld.config.xml file to configure the recognizer. As a matter of fact, if you want to play around with grammars, you should just be able to modify the hello.gram file and experiment to your heart's content. Just make sure you use words that are in the pronunciation lexicon. You can check for words by looking at the left hand column of models/acoustic/wsj/dict/cmudict.0.6d.
If you want to work from *.wav files, then you could modify HelloWorld.java to look more like HelloDigits.java.
Hope this helps,
Will
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I want to do some tests on the plateform which is like wavfile demo. my demo name is hopewavdemo."ant" is successful.But,when execute hopewavfile.jar ,the result is "blank" as follows:
E:\SPHINX\Sphinx4>java -jar bin\hopewavfile.jar
Loading Recognizer...
Decoding file:/E:/SPHINX/Sphinx4/bin/hopewavfile.jar!/demo/sphinx/hopewavfile/IH
.wav
WAVE (.wav) file, byte length: 337978, data format: PCM_SIGNED, 16000.0 Hz, 16 b
it, mono, little-endian, audio data, frame length: 168960
RESULT:
So,how can i solved this problems?any suggestion is appreciated!
Maggie You
Hi Will:
Thank you.However,because my poor English,I do not express that clearly.Now,I will say more.
I have done one demo named "hope"which is liked "HelloDigits"demo.The acoustic model was also trained by myself.I do not use Sphinx4's demos and models.
The "hope" demo worked well.But It demands that the "input files" should be from microphone.So, I want the input file is wavfile which is like "wavfile"demo.
I do one demo named "hopewavfile",but, the results is not what I want to get.I do not konw where the problem have occured.
Hope you could understand what I want to express. Thanks a lot in advance.
Maggie
Hi Maggie You:
There could be any number of problems here. I'm going to guess that your lH.wav file does not contain someone speaking just digits, but that it contains other words, and that your config file and grammar are still set up for digits.
However, the more important thing here is to understand what task/problem you are trying to support. If you describe this, we might be able to point you down another path.
Will
Hi Will:
Thanks a lot for your advice.I will say the task in more detail.
Shortly,I want to do one demo like wavfile.This demo's name is hopewavfile,which contains similar files with wavfile.Such as config.xml ,hopewavfile,hopewavfile.Manifest,hopewavfile.gram，IH.wav
hopewavfile.gram is:
JSGF V1.0;
/*
* JSGF Digits Grammar for Hello World example
/
grammar hope;
public <voice> = (EH | IH | UH | YU ) * ;
The other files are also modified.
The results is BLANK...
I do not konw how to solve it.
Maggie You
Hi:
There are two main problems here:
1) You are using the digits models (i.e., things really good at recognizing only numbers) with a grammar containing no digits.
2) Even if we get the config file set up to use more general models, at least one of the words in the grammar, IH, is not in the pronunciation lexicon.
An alternative approach for you might be to start with the demo/sphinx/helloworld demo, using its helloworld.config.xml file to configure the recognizer. As a matter of fact, if you want to play around with grammars, you should just be able to modify the hello.gram file and experiment to your heart's content. Just make sure you use words that are in the pronunciation lexicon. You can check for words by looking at the left hand column of models/acoustic/wsj/dict/cmudict.0.6d.
If you want to work from *.wav files, then you could modify HelloWorld.java to look more like HelloDigits.java.
Hope this helps,
Will