0

As the gram file that is used to convert audio to text, we have only a limited number of words that can be recognized. is there any way to transcribe any audio to text. Can we enhance the gram file to contain all the English words so that any of the English words can be recognized? Or can we do something else so that my purpose to convert any audio file to text could be released? I have wasted much time. Any idea will be appreciated. Thanks in advance.

Nikolay Shmyrev
  • 24,897
  • 5
  • 43
  • 87
Vishwanath
  • 149
  • 2
  • 14

1 Answers1

1

To get familar with speech recognition concepts read the tutorial first

http://cmusphinx.sourceforge.net/wiki/tutorial

Grammar that contains all the words is called a language model. The language model for US English is avialable for download here:

https://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20Generic%20Language%20Model/

To create language models for other languages you can check the tutorial

http://cmusphinx.sourceforge.net/wiki/tutoriallm

To decode with language model you can use Transcriber demo in sphinx4 or pocketsphinx_continuous binary with -lm option. Like this:

 pocketsphinx_continuous -infile file.wav -lm en_us.lm.dmp
Nikolay Shmyrev
  • 24,897
  • 5
  • 43
  • 87
  • Thanxx Nikolay.....can u please guide me how sphinx4.jar can be edited so that it can use the above model and takes an audio file as its input. – Vishwanath Feb 16 '14 at 15:11
  • In sphinx4 this model is already used by default in Transcriber demo. See for details http://cmusphinx.sourceforge.net/wiki/tutorialsphinx4. – Nikolay Shmyrev Feb 16 '14 at 15:17
  • but the hello world program included in sphinx4 uses a very short gram file. if it uses the language model , then what is the use of the gram file. and if it is needed then how can we increase the words in the gram file to detect any word spoken in an audio file. please help – Vishwanath Feb 16 '14 at 15:21
  • Please follow the link and read it carefully. It doesn't seem like you read it. – Nikolay Shmyrev Feb 16 '14 at 15:24
  • I read all the stuff....can you please help me how to use the language model you provided in sphinx4. the tutorial is a bit confusing to follow up in making a code that can convert audio to text.. please – Vishwanath Feb 16 '14 at 16:11
  • You need to checkout latest sources with subversion and check Transcriber demo. – Nikolay Shmyrev Feb 16 '14 at 17:24
  • hii Nikolay...as you told I checked all the things...can you please tell me how can I extract a language model (.lm) so as to replace the existing .trigram.lm file with the new one. – Vishwanath Feb 17 '14 at 18:35
  • Please tell hoe to use the en_us.lm.dmp file you provided in the transcriber demo. – Vishwanath Feb 17 '14 at 18:37
  • Hii Nicolay....please tell me how to use en_us.lm.dmp file you provided in the transcriber demo. Thank you so much in advance. – Vishwanath Feb 18 '14 at 07:30