I'm a beginner with sphinx3, I've download the sphinx3-0.6.3 release, compiled it.
I've run the regression test, it seems to wotk.
I'm kinf of lost now...
I guess I need to run the binary sphinx3_decode. But there are so many arguments, I can't make it work, there always are errors.
Could you please try to give me the right inputs argument, knowing that I'm willing to create a free speech recognizer (The user can say wathever he wants). I plan on using rm1 vocabulary and trigram langage model. I'm using raw files to check the decoder.
I realy hopes that someone will have the time to answer me. Any help would be very appriciate.
Best regards.
Cyril
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Welcome to the world of Sphinx! While I don't claim to be an expert on the subject, I've been playing with the engine for almost a year now. Here are some starting points for you:
Second, you will want to read extensively on SphinxTrain. While I admire your project goals, I'll have to say that if you are planning on creating an 'N' speech recognizer (nearly infinite speech possibilities), RM1 will not cater to that. You might want to start looking at the HUB4 open source acoustic models. HUB4 however, will probably only get you so far. Your best bet is to learn how to formulate linguistic models, and acoustic models yourself. While this takes time, you'll find that your results will probably be better.
Third, remember that patience will be your greatest virtue with Sphinx. Rome wasn't built in a day, and neither will your project.
Good Luck!
Shawn
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi guys,
I'm a beginner with sphinx3, I've download the sphinx3-0.6.3 release, compiled it.
I've run the regression test, it seems to wotk.
I'm kinf of lost now...
I guess I need to run the binary sphinx3_decode. But there are so many arguments, I can't make it work, there always are errors.
Could you please try to give me the right inputs argument, knowing that I'm willing to create a free speech recognizer (The user can say wathever he wants). I plan on using rm1 vocabulary and trigram langage model. I'm using raw files to check the decoder.
I realy hopes that someone will have the time to answer me. Any help would be very appriciate.
Best regards.
Cyril
Hi Cyril,
Welcome to the world of Sphinx! While I don't claim to be an expert on the subject, I've been playing with the engine for almost a year now. Here are some starting points for you:
First, read this manual:
http://www-2.cs.cmu.edu/~archan/documentation/sphinxDocDraft3.pdf
This is the Sphinx3 draft documentation. Granted, it's a draft - but there is a lot of useful information in there, including the arguments you are seeking.
Second, you will want to read extensively on SphinxTrain. While I admire your project goals, I'll have to say that if you are planning on creating an 'N' speech recognizer (nearly infinite speech possibilities), RM1 will not cater to that. You might want to start looking at the HUB4 open source acoustic models. HUB4 however, will probably only get you so far. Your best bet is to learn how to formulate linguistic models, and acoustic models yourself. While this takes time, you'll find that your results will probably be better.
Third, remember that patience will be your greatest virtue with Sphinx. Rome wasn't built in a day, and neither will your project.
Good Luck!
Shawn