I'd like to take my study-notes (verbally) and have them transcribed into text as I speak.
As I previously had a very good experience with Carnegie Mellon's Sikuli test tool, I believe that CMU Sphinx is a very good candidate for this application.
Does anyone know of a USB microphone that will work on the Raspberry PI (Raspian)?
And is CMU Sphinx a good speech-to-text engine for this application (of 'speaking' my study notes and having them converted into text as I speak).
TIA, Matt
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Raspberry PI is too slow for large vocabulary continuous speech recognition, I recommend you to look for more powerful hardware. Something Intel-powered probably.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2015-11-05
Nickolay, thanks for the reply. I'll go ahead and utilize my other desktop running, yep, Windoze.
I know the Raspberry Pi's not supposed to be used as a desktop. It's just that I've always had that teensy masochistic streak....
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Nickolay, would it not be possible to use pocketsphinx on the Pi2?
It is possible but for very simple task like keyword spotting or recognition of few commands from grammar. Large vocabulary requires careful tuning and decoding speed optimizaiton.
As far as I understand Jasper uses simple keyword spotting.
Additionally Matt, the Pine64 may be a better candidate for this than the Pi.
From the description it indeed looks better, I think it might be an option. It still needs some testing, it is hard to predict actual performance, you need to try it on the device.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I'd like to take my study-notes (verbally) and have them transcribed into text as I speak.
As I previously had a very good experience with Carnegie Mellon's Sikuli test tool, I believe that CMU Sphinx is a very good candidate for this application.
Does anyone know of a USB microphone that will work on the Raspberry PI (Raspian)?
And is CMU Sphinx a good speech-to-text engine for this application (of 'speaking' my study notes and having them converted into text as I speak).
Hello Matt
Raspberry PI is too slow for large vocabulary continuous speech recognition, I recommend you to look for more powerful hardware. Something Intel-powered probably.
Nickolay, thanks for the reply. I'll go ahead and utilize my other desktop running, yep, Windoze.
I know the Raspberry Pi's not supposed to be used as a desktop. It's just that I've always had that teensy masochistic streak....
Nickolay, would it not be possible to use pocketsphinx on the Pi2?
The Jasper Project seem to be using it (they're using additional tools, though).
I'm currently attempting to get this to work on the Pi2, accuracy is pretty bad, probably because of the speed but I might be able to optimize that.
Additionally Matt, the Pine64 may be a better candidate for this than the Pi.
It is possible but for very simple task like keyword spotting or recognition of few commands from grammar. Large vocabulary requires careful tuning and decoding speed optimizaiton.
As far as I understand Jasper uses simple keyword spotting.
From the description it indeed looks better, I think it might be an option. It still needs some testing, it is hard to predict actual performance, you need to try it on the device.
You're right about Jasper using simple keywords. Apparently the jasper-dev branch can recognize sentences but I haven't had the chance to test it yet.
Either way, CMU Sphinx looks like the way to go. Glad I came across this project.