Closed ghost closed 7 years ago
In the latest version on github, you have to manually set the threshold depending on your ambient mic/noise level. To do this, you can edit the "athena/keyphrases.txt" file and change the values: athena /1e-30/ to something like athena /1e-20/ or athena /1e-10/ (less sensitive) (or up to around /1e-60/ if you need it to detect more easily)
The wake-up-word/keyphrases use pocketsphinx behind the scenes. I'm hoping to find a way to auto-adjust the thresholds soon (e.g. have a background thread adjust for the ambient noise level... then if false triggers still occur, the user can respond saying "false trigger" and athena will adjust to a lower sensitivity threshold"
I see. I've just been editing the modules in my dist-packages directory lol. Thank you for answering!
Not even speaking, too. Just noises.
This doesn't happen on the distribution version, but when I run it from source, the issue happens.