I've create acoustic models via SphinxTrain and now I am with the "How to Use Models from SphinxTrain in Sphinx-4" tutorial. My problem is that I don't know which files I have to copy to the sphinx4's acoustic folder, since when I run the training process a lot of files were created (much more than it's specified in the "How to Use Models from SphinxTrain in Sphinx-4" tutorial). So, are they all the files contained in "model_architecture" and the "model_parameters" folder ?
Plus, my application is supposed to be a word-spotting application. I would like to know if that would be possible, since I've read that it's no support for word-spotting right now, in the FAQ.
Thanks !!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
> Plus, my application is supposed to be a word-spotting application. I would like to know if that would be possible, since I've read that it's no support for word-spotting right now, in the FAQ.
about word spotting, this topic was discussed recently, look on the topics list.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
or in the similar folder according to your number of senones.
you need mdef, mixture_weights, transition_matrices and variances. In older sphinxtrain mdef was located in model_architecture folder and had 1000_8 suffix as well.
also you need a dictionary and a filler dictionary, look into existing models in sphinx4/models for example.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I've create acoustic models via SphinxTrain and now I am with the "How to Use Models from SphinxTrain in Sphinx-4" tutorial. My problem is that I don't know which files I have to copy to the sphinx4's acoustic folder, since when I run the training process a lot of files were created (much more than it's specified in the "How to Use Models from SphinxTrain in Sphinx-4" tutorial). So, are they all the files contained in "model_architecture" and the "model_parameters" folder ?
Plus, my application is supposed to be a word-spotting application. I would like to know if that would be possible, since I've read that it's no support for word-spotting right now, in the FAQ.
Thanks !!
> Plus, my application is supposed to be a word-spotting application. I would like to know if that would be possible, since I've read that it's no support for word-spotting right now, in the FAQ.
about word spotting, this topic was discussed recently, look on the topics list.
Thanks, but I still have some doubts. I've opened a new post where I try to explain my doubts clearly.
Thank you !!
Usually the whole model is located in
model_parameters/your_name.cd_cont_1000_8
or in the similar folder according to your number of senones.
you need mdef, mixture_weights, transition_matrices and variances. In older sphinxtrain mdef was located in model_architecture folder and had 1000_8 suffix as well.
also you need a dictionary and a filler dictionary, look into existing models in sphinx4/models for example.