I would like to apologize if this is a trivial question, I have tried googling and reading the tutorials but I'm afraid I haven't been able to find what I'm looking for. I am quite new to CMU Sphinx and speech recognition in general although I've worked on some sound recognition projects in the past.
The short story: I\m trying to implement a limited vocabulary isolated word recognition of certain Arabic words, using PocketSphinx for Android. I have found a trained Arabic acoustic model at the link below:
My idea was to take that existing acoustic model and provide my own simple dictionary and a simple grammar. Unfortunately, the results I'm getting are quite disappointing, even the English model provided for the demo PocketSphinx for Android application (en-us-ptm) does a better job.
My question revolves around which files exactly to use as my acoustic model from that link above?
I'm taling about this file specifically (Trained model parameters): http://faculty.kfupm.edu.sa/SE/elshafei/AASR_Model.rar
From what I can figure out, there are some intermediate folders in that RAR archive, and I'd like to take the final AM, but unfortunately I know very little about the format of the SphinxTrain output beyond the basics. Could someone help me out here with pointing out which files constitute the final model, or by pointing me in the right direction to documentaiton where I'd be able to learn that myself?
If that acoustic model is even applicable for PocketSphinx (I've read something about unsupported features and such), but that's another pair of shoes, I suppose.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
I would like to apologize if this is a trivial question, I have tried googling and reading the tutorials but I'm afraid I haven't been able to find what I'm looking for. I am quite new to CMU Sphinx and speech recognition in general although I've worked on some sound recognition projects in the past.
The short story: I\m trying to implement a limited vocabulary isolated word recognition of certain Arabic words, using PocketSphinx for Android. I have found a trained Arabic acoustic model at the link below:
http://faculty.kfupm.edu.sa/SE/elshafei/AASR.htm#%28A%29
My idea was to take that existing acoustic model and provide my own simple dictionary and a simple grammar. Unfortunately, the results I'm getting are quite disappointing, even the English model provided for the demo PocketSphinx for Android application (en-us-ptm) does a better job.
My question revolves around which files exactly to use as my acoustic model from that link above?
I'm taling about this file specifically (Trained model parameters): http://faculty.kfupm.edu.sa/SE/elshafei/AASR_Model.rar
From what I can figure out, there are some intermediate folders in that RAR archive, and I'd like to take the final AM, but unfortunately I know very little about the format of the SphinxTrain output beyond the basics. Could someone help me out here with pointing out which files constitute the final model, or by pointing me in the right direction to documentaiton where I'd be able to learn that myself?
If that acoustic model is even applicable for PocketSphinx (I've read something about unsupported features and such), but that's another pair of shoes, I suppose.
This model you are trying to use is very basic and trained with the very old sphinxtrain. It is not going to work on Android
You have to train a model yourself.
Thanks a lot Nickolay!
Last edit: Denis Gerina 2015-08-29