I am trying to achieve a high match speech recognition or no match at all.
I managed to setup pocket-sphinx with Android.
My app let users input the restaurant of their menu.
I have a dictionary with ~3000 words of the most common dishes of a restaurant's menu.
I intend to use sphinx as a support input method.
EXAMPLE
the user must enter the word peperoni inside on a textfield.
He can either input the word using the on-screen touch keyboard or he speaks the word 'peperoni'.
Sphinx listens continuously in the background when the textfield is selected.
A clickable toast message should be shown only if a very close match to a word in the dictionary is found.
In other words I don't need sphinx to make the best case even if it's pretty far.
I need sphinx to notify of matches only when they are pretty close or to not make any guess at all.
Any suggestion on how to achieve this?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It is not easy to get it listen for 3000 words at once and I doubt it is required. If you want to detect just 10-20 words you can use keyword spotting mode. It is demonstrated in android demo, see also here:
Hi there,
I am trying to achieve a high match speech recognition or no match at all.
I managed to setup pocket-sphinx with Android.
My app let users input the restaurant of their menu.
I have a dictionary with ~3000 words of the most common dishes of a restaurant's menu.
I intend to use sphinx as a support input method.
EXAMPLE
the user must enter the word peperoni inside on a textfield.
He can either input the word using the on-screen touch keyboard or he speaks the word 'peperoni'.
Sphinx listens continuously in the background when the textfield is selected.
A clickable toast message should be shown only if a very close match to a word in the dictionary is found.
In other words I don't need sphinx to make the best case even if it's pretty far.
I need sphinx to notify of matches only when they are pretty close or to not make any guess at all.
Any suggestion on how to achieve this?
It is not easy to get it listen for 3000 words at once and I doubt it is required. If you want to detect just 10-20 words you can use keyword spotting mode. It is demonstrated in android demo, see also here:
http://stackoverflow.com/questions/25748113/recognizing-multiple-keywords-using-pocketsphinx
So most likely you will have to restrict the keyword list to the actual dishes of that particular restaurant.