I have a medium sized vocabulary with around 250 words.
I was confused on which type of acoustic modeling to use, connected word model or tri-phone model.
Which would be more efficient and accurate ?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
By "connected word model" I mean connected word recognition problem.
I read that for small/medium vocablary instead of using HMM-GMM model at sub-word units we can use level building / Two level DP algorithm based on individual word model.
Just wanted to check which would be better in my case of speaker independent recognition for a 250 word vocablary.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I have a medium sized vocabulary with around 250 words.
I was confused on which type of acoustic modeling to use, connected word model or tri-phone model.
Which would be more efficient and accurate ?
What do you mean by "connected word model".
I think I quoted it incorrectly.
By "connected word model" I mean connected word recognition problem.
I read that for small/medium vocablary instead of using HMM-GMM model at sub-word units we can use level building / Two level DP algorithm based on individual word model.
Just wanted to check which would be better in my case of speaker independent recognition for a 250 word vocablary.
You already asked your question here:
https://sourceforge.net/p/cmusphinx/discussion/speech-recognition/thread/bf4e8962/#61c3
there is no need to ask the same thing many times