voice detection

there are a couple things i would like to have my Linux system do for me that deals with the microphone and voice technology.

1. while i am asleep i will have Linux be my alarm clock (already did this much). i want to add limited means to tell it by voice to shut off the alarm (execute my "alarm off" script). i do not need a full voice to text capability. i would prefer it be as accurate as possible for a few basic phrases like "alarm off". especially important is when my "alarm off" script plays the voice "the alarm is off" it won't trigger the voice detector in any way. likewise no other words can do this. if someone else says the same phrase, it's fine to do the same thing.

2. while i am away or asleep i want to ha any voice (anyone) saying anything in any language start recording of sound including the sound a few seconds or so before that voice as well as executing a script. that means the program will be continuously be keeping a looping buffer of sound and dump that out to start the recording file. then it would stop after 4 minutes of no voice being detected and go back to listen-only mode.

how much voice technology would i need to get this going?

One of the projects to look at for that would be Mycroft.

Originally Posted by Skaperen how much voice technology would i need to get this going?

I think about 100MB.

Seriously though, I saw a video (years ago now) where someone wrote code with his voice, on a Linux machine. Some python stuff and a normal microphone IIRC.

i want to detect voice among other noises and trigger a process only if there was a voice ... specific words for one project ... any words in any language for another. singing counts as voice. can general voice recognition do that? what if someone speaks Japanese?

Originally Posted by Skaperen i want to detect voice among other noises and trigger a process only if there was a voice ... specific words for one project ... any words in any language for another. singing counts as voice. can general voice recognition do that? what if someone speaks Japanese?

Are the projects you checked out meanwhile capable to do that? If yes, please do share, I'm genuinely interested.

so far, i am unable to make that determination. Mycroft appears to be a different kind of project, more like Alexa or Siri. it requires registration at their home site to work, something i would not do for my projects. others i have found just don't go into enough detail to really know.