I have read through the CMU documentations regarding the different versions of
recognizer. As per my understanding gathered from the documents Sphinx2 is the
fast recognizer of all, but the accuracy is the least one. What is the diff.
between Sphinx4 /Sphinx3 when it comes to performance? We are planning to use
the recognizer for IVR which requires real-time recognition. Which is the
recognizer to be considered?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The performance of the recognizer largely depends on the acoustic model, not
the recognizer itself. From that point of view all recognizers are more or
less equal
You need to us select recognizer based on other criteria. See
I went through the decoders list that you had sent. PocketSphinx is suggested
to be used for desktop,mobile applications. Whereas Sphinx2 is the obsolete
decoder. It means there are only 2 versions available to be used for IVR.
Sphinx3/Sphinx4. Is that right?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
But the link mentions that pocketsphinx is to be used for desktop/mobile
apps not for IVR.
I do not see a contradiction here. The fact pocketsphinx is used for mobile
apps doesn't mean it can not be used for IVR.
It's already actively used in UniMRCP project I linked you above in thousands
installations.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi All,
I have read through the CMU documentations regarding the different versions of
recognizer. As per my understanding gathered from the documents Sphinx2 is the
fast recognizer of all, but the accuracy is the least one. What is the diff.
between Sphinx4 /Sphinx3 when it comes to performance? We are planning to use
the recognizer for IVR which requires real-time recognition. Which is the
recognizer to be considered?
The performance of the recognizer largely depends on the acoustic model, not
the recognizer itself. From that point of view all recognizers are more or
less equal
You need to us select recognizer based on other criteria. See
http://cmusphinx.sourceforge.net/wiki/versions
If you need high performance, you can choose pocketsphinx. See UniMRCP
http://www.unimrcp.org/ and pocketsphinx unirmcp
plugin for details
For more flexibility like web Java EE integration use Sphinx4. See OpenIVR
http://www.spokentech.org for details.
I went through the decoders list that you had sent. PocketSphinx is suggested
to be used for desktop,mobile applications. Whereas Sphinx2 is the obsolete
decoder. It means there are only 2 versions available to be used for IVR.
Sphinx3/Sphinx4. Is that right?
No. As I wrote above you can use either pocketsphinx or sphinx4.
But the link mentions that pocketsphinx is to be used for desktop/mobile apps
not for IVR.
I do not see a contradiction here. The fact pocketsphinx is used for mobile
apps doesn't mean it can not be used for IVR.
It's already actively used in UniMRCP project I linked you above in thousands
installations.