5

I was reading this guide on speech recognition, and it mentioned that I need three items for speech recognition: Acoustic model, Language Model, Phonetic Dictionary.

I wanted to start playing with this python demo, which uses Gstreamer to capture from the mic and resample to 8kHz, 16-bit PCM audio.

I see that I can specify the language model and phonetic dictionary, and I use the one [provided by cmu]:

http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20HUB4%20Language%20Model/

But I am confused where I should specify the acoustic model? Does gstreamer have its own acoustic model I'm implicitly using? I was hoping to use the acoustic model provided here for slightly better results:

 http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20HUB4%20Acoustic%20Model/

(Sorry about the hyperlinks. I can't post more than 2 links with rep less than 10)

user1812844
  • 297
  • 2
  • 10

1 Answers1

3

But I am confused where I should specify the acoustic model?

You can specify the model with the hmm property of the gstreamer element. Just like it's covered in tutorial

asr.set_property('lm', '/home/user/mylanguagemodel.lm')

You can use

asr.set_property('hmm', '/home/user/acoustic_model_folder')

Does gstreamer have its own acoustic model I'm implicitly using?

Yes, by default it uses US English model hub4wsj_sc_8k from the distribution

Nikolay Shmyrev
  • 24,897
  • 5
  • 43
  • 87