Hi, I am interested in developing a Chinese Accoustic Model.  How can I go about doing it? 

Can I use the text-to-speech software to do it ?

Thanks,

Mark

--- (Edited on 7/6/2008 3:30 am [GMT-0500] by Visitor) ---

Hi Mark,

>interested in developing a Chinese Accoustic Model.  How can I go about doing it? 

Follow the VoxForge Tutorial to get a good understanding of how to create an acoustic model in English.

Then, modifying the VoxForge Tutorial where appropriate, create a simple Chinese acoustic model as follows (these are the high-level steps, see the VoxForge Tutorial for details):

• create a list of sentences (training set);
• record yourself reading each of these sentences;
• Create a Chinese phone set (list of the phones used in your dialect of Chinese); VoxForge English phoneset is derived from the CMU phone set;
• create a pronunciation dictionary with entries for each word in your training set; VoxForge uses the CMU pronunciation dictionary for its English language acoustic models;
• generate acoustic models using the process described in the VoxForge Tutorial.

This will allow you to create monophone acoustic models (up to step 8). 

To create tied-state triphone acoustic models, you will need to create 'questions' (see the tree.hed script in step 10).  I just used the one included with the HTK toolkit, and am not familiar with creating one for another language.  Sphinx has an algorithm to automatically create the questions.

See my post on this thread for a bit more information.

Ken

--- (Edited on 7/12/2008 12:31 am [GMT-0400] by kmaclean) ---