Acoustic Model Discussions

Flat
Re: Speech Recognition Model Converter
User: dara
Date: 1/19/2011 9:06 am
Views: 95
Rating: 7

Hi,

1. You don't need to create both files beforehand, the application will make them for you based on whatever names you like.. :)

2.That error usually happens because of storing something whose size exceeds the storage capacity, which is "htkphone" in your case. What you could do is to edit the size of "htkphone" (now, its size is 10, change it to 20 for example) in "src/programs/s3mod2htk/s3mod2htk.c". Then try to buid it again.

3. You need to create the tiedlist by hand. It tells HTK the classification of each phone of your phone set, e.g. which phones are nasal, which ones are vowels, and so on. You can have a look at IPA (International Phonetic Alphabet) as a guidance on how to categorize your phone set.

 

regards,

Dara

--- (Edited on 1/19/2011 9:06 am [GMT-0600] by dara) ---

Re: Speech Recognition Model Converter
User: enigma1987
Date: 1/20/2011 3:56 am
Views: 99
Rating: 7

Hi,

Thanks I successfully compiled and use s3mod2htk both in windows and linux.

Another question that I want to ask is can we use s3mod2htk to convert semi-continuous acoustic models (pocketsphinx models) to HTK models ?

I have tried to convert a semi-cont. acoustic model which is about 10 mb. , it converted with no error but the output model htkmmf is approximately 6 gb. I guess something was wrong.

 

Thanks a lot for your help,

Berker.

 

--- (Edited on 1/20/2011 3:56 am [GMT-0600] by enigma1987) ---

Re: Speech Recognition Model Converter
User: dara
Date: 1/25/2011 11:00 am
Views: 334
Rating: 4

Hi,


I guess it can convert the pocketsphinx model to HTK format, but I'm not sure, to be honest. Maybe you can have a look at the content of the converted model, to see if there's something wrong or odd although what's wrong and odd is not obvious (if there's any).

Sorry buddy, can't help you for that.

 

regards,

Dara

--- (Edited on 1/25/2011 11:00 am [GMT-0600] by dara) ---

Re: Speech Recognition Model Converter
User: esuncloud
Date: 1/3/2012 2:04 pm
Views: 327
Rating: 7

Hi, I am trying to convert the acoustic model using this tool too.

It seems that there will be some errors if you use vs2008, and it will be fine to compile with lower version vc6.0.

However, I have the same problem now, i.e. it will output a huge file from a small Sphinx model? Have you solved this problem? Thanks.

--- (Edited on 1/3/2012 2:04 pm [GMT-0600] by esuncloud) ---

Re: Speech Recognition Model Converter
User: Abhilash
Date: 6/27/2012 10:25 am
Views: 1948
Rating: 5

Hi Berker,

I too got the exe file s3mod2htk. But how can I give the parameters and inputs files to the exe file. I'm not familiar with visual  C++ 2008.  I have the sphinx acoustic model files with me. I am not able to generate the output files htkmmffn and htkmacrofn. Please help me. Thanks for your help.

Abhilash

 

--- (Edited on 6/27/2012 10:25 am [GMT-0500] by Visitor) ---

PreviousNext