The KLearnNotes2 Handbook | ||
---|---|---|
<<< Previous | Next >>> |
You can use your microphone to provide answers to KLearnNotes2 tests, and even to give commands to the program. To use this feature you need to do some setup though. I expect it to take (including reading the instruction bellow) less then 5 minutes for a novice, unexperienced user to make first voice-answered test. :) See "Quick Start" section below.
It may still take some time to fine-tune microphone settings and voice model to use voice recognition really smoothly. Therefore, I recommend, that once you do basic setup you should read details in next sections.
Setup consists of two steps: microphone calibration and preparing speaker voice model (i.e. recording samples).
You do microphone calibration by selecting Menu Options -> Voice Input -> Microphone setup. This is a step-by-step wizard which will guide you through the process. Just click on appropriate buttons, speak when you are asked to speak, be silent when you are asked to. :)
Note, that on some soundcards mic setup does not recognize igain. You can still fine-tune this with some other programs (e.g. kmix). Hint: igain values about 10% are reasonable.
Choose Menu Options->Voice Input->Speaker Model Setup. If you never run this before, a new voice model (speaker model) will be automatically started. Click on 'Notes' tab. On the left you see list of notes.
click on one of note's names, e.g. "C"
click on 'Record a sample' and say "see" (or whatever is the sound for letter "C" in your language; Poles and Germans would say "tzae") to your microphone - this should add one item to the right list - list of samples for note "C"; the sample will be played immediately after recording - please, listen if it really sounds like "C" (for example if the beginning/ending is not cut off)
repeat recording three more times, until the label by the letter "C" on the left list changes from "4 lacking" to "(OK)"
Note, that although 4 is the minimum number of samples, voice recognition will be more correct if you record some additional samples (say, up to 8 samples total for one note/command).
I suggest that, just to check if everything works, you record samples only for two notes, say "C" and "G", and save such a simple voice model. You can always add samples for other notes in future!
When you are done, you need to save the model to a file (click on 'Save as' button, and choose a name, for example "C_and_G-your-name").
When you exit voice model editor (described in the last section)
by clicking OK, the voice model you have just saved will be auto selected
for recognition process.
Now, to start recognition process just click on 'Voice input' button (the
one with a microphone
) in the toolbar.
Say "see" to your microphone. This should highlight the "C" namebutton and all "C" notes. If you say "gee" the same should happen with namebutton "G". Now, you can select Exercise 01 (which has only C and G notes), and start test. You should be able to give answers by your voice! :)
CVoiceControl should save microphone settings. Well, in my case it doesn't recognize igain, and doesn't save it.
Simple remedy: each time I start KLearnNotes2 with intention of voice recognition, I start kmix (or other mixer), and adjust i-gain level manually (to about 10%).
Microphone config saves settings to ~/.cvoicecontrol/config file.
Voice model filename is chosen by a user. Suggested directory is ~/.klearnnotes2_d/voice_models/.
On exit, KLearnNotes2 will save last voice model filename used in ~/.klearnnotes2_d/current_voice_model file and will use it to start voice recognition when run again.
<<< Previous | Home | Next >>> |
It's not a bug, it's a feature! - FAQs | Voice input - General description |