VoxClick project - control mouseclicks by voice
Posted: Fri 18 Oct 2013, 20:07
UPDATED Oct 27 2013
(minor update. Addition of troubleshooting/solution info in case of window resize/move freeze)
Wouldn't it be nice to avoid the mechanical difficulty of handling clicks and doubleclicks on your mouse? What if you could TELL the computer when to do a click, doubleclick, rightclick etc?
You would position the mouse exactly where you want it then just give the computer a spoken command eg: "RIGHTCLICK"
The voxclick004.pet 22MByte download is available here:
http://www.mediafire.com/download/5ybrn ... ick004.pet
(If you have already installed the audiotest .pet below, please uninstall it first or at least move any previous voxclick related scripts out of the /root/Startup directory)
The new .pet installs voxclick into the /root/Startup directory so after installing the new .pet just restartx (menu,shutdown,restart x server) and it will be active and ready to receive your voice commands. No voice training is required. Also, a voxclick004 howto document gets loaded into /root.
The following commands are recognised:
CLICK
RIGHTCLICK
CENTERCLICK
DOUBLECLICK
CLICKHOLD
RELEASE
I have also attached a voxclick howto document to this post.
If you are not ready to try the complete .pet yet and just want to test your audio I have also have uploaded a .pet which allows the user to test the quality of voice recognition on their system - without any risk of actual commands being activated. This is a very easy test and requires no voice training etc - just install the pet and read the "how to start" document which you will find in /root.
voxclick_and_sphinx_audiotest.pet 22MByte download here:
http://www.mediafire.com/download/xn5bb ... iotest.pet
Within a few minutes this audiotest will let you see if this project will work for you.
Troubleshooting :
1) In most cases any problems are caused by mic and capture volume being TOO LOUD rather than too quiet. On some machines you may have to turn these settings down so far that you would think the system could not hear the mic - but sometimes that is the right setting...
(and you may have to turn OFF the +20db mic boost)
2) Some puppies exhibit a symptom where window dragging or window resizing does not allow the cursor to "release" so you can't let go of the window corner or whatever. This can be cured by doing the following:
Use geany to open file /root/.jwm/jwmrc-personal as text (note the dot - .jwm is hidden). Part way down the file you will find two parameters called "move mode" and "resize mode". It is necessary to change each of these from 'outline' to 'opaque'.
3) When pocketsphinx first runs and starts listening it seems to be in a mode where it wants to auto adjust it's sensitivity (or at least that is what I interpret...) and it might get the first word incorrect before it "locks in" on the correct settings. I find it useful to begin with a loudly spoken sentence that is not in the special vocab list. Something like "the rain in spain falls mainly on the plain". That seems to wake it up and from there on it recognises the normal command words spoken at normal volume. Might depend on hardware; I don't know, but worth a try anyway if you are having problems.
.
(minor update. Addition of troubleshooting/solution info in case of window resize/move freeze)
Wouldn't it be nice to avoid the mechanical difficulty of handling clicks and doubleclicks on your mouse? What if you could TELL the computer when to do a click, doubleclick, rightclick etc?
You would position the mouse exactly where you want it then just give the computer a spoken command eg: "RIGHTCLICK"
The voxclick004.pet 22MByte download is available here:
http://www.mediafire.com/download/5ybrn ... ick004.pet
(If you have already installed the audiotest .pet below, please uninstall it first or at least move any previous voxclick related scripts out of the /root/Startup directory)
The new .pet installs voxclick into the /root/Startup directory so after installing the new .pet just restartx (menu,shutdown,restart x server) and it will be active and ready to receive your voice commands. No voice training is required. Also, a voxclick004 howto document gets loaded into /root.
The following commands are recognised:
CLICK
RIGHTCLICK
CENTERCLICK
DOUBLECLICK
CLICKHOLD
RELEASE
I have also attached a voxclick howto document to this post.
If you are not ready to try the complete .pet yet and just want to test your audio I have also have uploaded a .pet which allows the user to test the quality of voice recognition on their system - without any risk of actual commands being activated. This is a very easy test and requires no voice training etc - just install the pet and read the "how to start" document which you will find in /root.
voxclick_and_sphinx_audiotest.pet 22MByte download here:
http://www.mediafire.com/download/xn5bb ... iotest.pet
Within a few minutes this audiotest will let you see if this project will work for you.
Troubleshooting :
1) In most cases any problems are caused by mic and capture volume being TOO LOUD rather than too quiet. On some machines you may have to turn these settings down so far that you would think the system could not hear the mic - but sometimes that is the right setting...
(and you may have to turn OFF the +20db mic boost)
2) Some puppies exhibit a symptom where window dragging or window resizing does not allow the cursor to "release" so you can't let go of the window corner or whatever. This can be cured by doing the following:
Use geany to open file /root/.jwm/jwmrc-personal as text (note the dot - .jwm is hidden). Part way down the file you will find two parameters called "move mode" and "resize mode". It is necessary to change each of these from 'outline' to 'opaque'.
3) When pocketsphinx first runs and starts listening it seems to be in a mode where it wants to auto adjust it's sensitivity (or at least that is what I interpret...) and it might get the first word incorrect before it "locks in" on the correct settings. I find it useful to begin with a loudly spoken sentence that is not in the special vocab list. Something like "the rain in spain falls mainly on the plain". That seems to wake it up and from there on it recognises the normal command words spoken at normal volume. Might depend on hardware; I don't know, but worth a try anyway if you are having problems.
.