mirror of
https://github.com/Ponce/slackbuilds
synced 2024-11-24 10:02:29 +01:00
50 lines
2.5 KiB
Text
50 lines
2.5 KiB
Text
|
You may want to install the Hidden Markov Model Toolkit (HTK) which is
|
||
|
|
covered by a license which does not permit free distribution. However,
|
||
|
|
you need HTK if you want to train your acoustic model. You can obtain
|
||
|
|
HTK from here (but only after registering): http://htk.eng.cam.ac.uk/
|
||
|
|
|
||
|
|
If you are creating solutions which will be used by more than one user, or
|
||
|
|
simply don't have the time to train the system, you can use static base models.
|
||
|
|
Static models are used as-is and are not modified by simon in any way.
|
||
|
|
Because of this, it is important that the selected base model matches your
|
||
|
|
voice as closely as possible.
|
||
|
|
|
||
|
|
Even if you use a static model, you NEED to get an acoustic model from the web.
|
||
|
|
You can download some prebuilt models at http://www.voxforge.org/
|
||
|
|
|
||
|
|
BEGINNER GUIDE:
|
||
|
|
If you are a beginner and you don't know exactly how a speech recognition works,
|
||
|
|
but want just to enable this "cool feature", you may want to follow these steps
|
||
|
|
(static model), in order to make simon operative (English).
|
||
|
|
|
||
|
|
This is to help you to your first approach to this program, next you will
|
||
|
|
be able to customize more and more!
|
||
|
|
|
||
|
|
0. Browse acoustic models from
|
||
|
|
http://www.repository.voxforge1.org/downloads/Nightly_Builds/current/
|
||
|
|
Download "HTK_AcousticModel-2010-12-16_16kHz_16bit_MFCC_O_D.tgz"
|
||
|
|
1. Uncompress the model where you want.
|
||
|
|
2. Run "ksimond" (not from root). You need to have the daemon simond running.
|
||
|
|
3. Configure "simond". (ksimond -> configuration -> simond). Add a username and a
|
||
|
|
password which are going to be used by simon.
|
||
|
|
4. Run "simon". An assistant will appear. Click "Next" once to jump to "Scenarios"
|
||
|
|
section of the assistant.
|
||
|
|
5. Get some scenario. You need at least one, download a scenario in English.
|
||
|
|
6. Configure base model. Choose "Static model" type. From the uncompressed acoustic
|
||
|
|
model of step 1 choose:
|
||
|
|
- "hmmdefs" file for HMM definition
|
||
|
|
- "tiedlist" file for Tiedlist
|
||
|
|
- "macros" file for Macros
|
||
|
|
- "stats" file for Stats
|
||
|
|
Click "Ok", then "Next".
|
||
|
|
7. "Server" and "Sound devices" sections configuration depends on what hardware
|
||
|
|
and software you're going to use.
|
||
|
|
You can safely just press "Next" to leave them unchanged.
|
||
|
|
8. Adjust the volume of your microphone (or any input device you're going to use)
|
||
|
|
I suggest you to get the rumor at few percentage (3%-4%) and to get
|
||
|
|
"Volume correct" while speaking (I boosted my microphone for that)
|
||
|
|
9. Optionally perform a training of speechable texts of your scenario to put your
|
||
|
|
voice in training data for a better recognition.
|
||
|
|
10. Speak!
|
||
|
|
|