I am trying to use Sphinx for building an automatic audio categorizer. My objective is to derive the subject or topic of an audio file by seeking some "magical words" in the audio using sphinx4. Im not looking for real time functionalities, indeed Im thinking in using several times the recognizer with different small dictionaries and later collect a collection of the results through some statistical model (yet not defined).
I would like to know which is the best sphinx configuration in order to achieving this goal.
I had thought in using the HUB4 language model, HUB4 acoustic model and a very limited dictionary (this dictionary will contain the "magical words"). Is this the best choice?
In my own tests I have discovered a huge WER increasing depending on the source sound quality. Where can I find good audio samples more complex than AN4.
Thanks a lot and best regards.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello
I am trying to use Sphinx for building an automatic audio categorizer. My objective is to derive the subject or topic of an audio file by seeking some "magical words" in the audio using sphinx4. Im not looking for real time functionalities, indeed Im thinking in using several times the recognizer with different small dictionaries and later collect a collection of the results through some statistical model (yet not defined).
I would like to know which is the best sphinx configuration in order to achieving this goal.
I had thought in using the HUB4 language model, HUB4 acoustic model and a very limited dictionary (this dictionary will contain the "magical words"). Is this the best choice?
In my own tests I have discovered a huge WER increasing depending on the source sound quality. Where can I find good audio samples more complex than AN4.
Thanks a lot and best regards.