General Discussion

Nested
Basic help needed please
User: jineshgeorge
Date: 3/26/2010 5:15 am
Views: 4036
Rating: 1

 

 

Hi,

As a part of my project, I need to recognize 'yes' and 'no' from the user through the microphone. I intend to use Julius for this. I followed these steps 

http://www.voxforge.org/home/dev/acousticmodels/linux/create/htkjulius/tutorial/data-prep

but I am not sure, how to create the '.grammar' and '.voca' file.  what about the prompts file, how do I create that, because I am just dealing with 2 words 'yes' and 'no'. Can anyone please help me regarding this?

Thanks in advance..

 

 

 

 

 

 

--- (Edited on 3/26/2010 5:15 am [GMT-0500] by ) ---

Re: Basic help needed please
User: kmaclean
Date: 3/29/2010 9:53 am
Views: 1707
Rating: 2

>how to create the '.grammar' and '.voca' file.

Find the pronunciation of the words you want in the lexicon and compile a grammar file as show in Step 1.  Something like:

sample.grammar

S : NS_B SENT NS_E
SENT: COMMAND

sample.voca

% NS_B
<s>        sil

% NS_E
</s>        sil

% CALL_V

YES    y eh s

NO       n ow

Record the all the prompts shown in Step 3.

Carry on with the other steps to get a basic understanding of the process.

Then, you can look at creating an acoustic model with just audio for "yes" and "no".

Ken

--- (Edited on 3/29/2010 10:53 am [GMT-0400] by kmaclean) ---

PreviousNext