I'm willing to develop a speech recognition application using PocketSphinx
(very simple, no fancy stuff).
The language is french
It must be speaker independent
The vocabulary is small (~100 words)
The words are not in the dictionary (they're brand names for the most part)
There's no such thing as a sentence, I just need to recognize one or two words)
It needs to run fast on an ARM device with good accuracy.
I don't really know where to start though, I have a list of all the words and
that's it.
Acoustic model: there are existing french acoustic models but they're quite
too big for my need, what should I do there? Create my own or use these ?
Dictionary: I have no idea how to create that, the wiki page gives a list of
tools but I don't really understand how they work. I considered using espeak
but have no idea how to do that.
I'm starting to wonder if CMU Sphinx might not be overkill for such a simple-
looking yet complicated task.
Thanks in advance for helping me, and thanks for your hard work on providing
people with great open source software!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Acoustic model: there are existing french acoustic models but they're quite
too big for my need, what should I do there? Create my own or use these ?
If you have time to create an acoustic model and required data (as listed in
acoustic model training tutorial) then you can create an acoustic model. It's
not an easy task.
Dictionary: I have no idea how to create that, the wiki page gives a list of
tools but I don't really understand how they work. I considered using espeak
but have no idea how to do that.
This is a least of your problems.
I recommend you to read the tutorial to become familar with pocketsphinx and
other tools. If you really want to accomplish your task you need to prepare to
some effort.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi!
I'm willing to develop a speech recognition application using PocketSphinx
(very simple, no fancy stuff).
I don't really know where to start though, I have a list of all the words and
that's it.
Acoustic model: there are existing french acoustic models but they're quite
too big for my need, what should I do there? Create my own or use these ?
Dictionary: I have no idea how to create that, the wiki page gives a list of
tools but I don't really understand how they work. I considered using espeak
but have no idea how to do that.
I'm starting to wonder if CMU Sphinx might not be overkill for such a simple-
looking yet complicated task.
Thanks in advance for helping me, and thanks for your hard work on providing
people with great open source software!
This task is easy to describe but not easy to implement
This counted like a medium vocabulary according to the tutorial
http://cmusphinx.sourceforge.net/wiki/tutorial
If you have time to create an acoustic model and required data (as listed in
acoustic model training tutorial) then you can create an acoustic model. It's
not an easy task.
This is a least of your problems.
I recommend you to read the tutorial to become familar with pocketsphinx and
other tools. If you really want to accomplish your task you need to prepare to
some effort.