parameter tuning to detect speech generally

[never_is]

Hey there,

I have a general question: What are the optimal parameter values, to be able to detect speech in general (if a person is (or several are) talking)? I don’t need to check for certain words, the only thing of interest to me, is whether somebody is speaking or not (I assume this works via frequencies?).
I’m not an advanced programmer, and i simply managed to integrate openEars into my swift project (recognition of words works fine).
But right now it is detecting speech even when I just tap on the desk (App runs on iPhone 6s, iOS 9.2.1).
How and in what way can I fine-tune it, such that it is able to reliably detect if somebody is actually speaking?

Thanks a lot for helping!
S.

[Halle Winkler]

Welcome,

There is no universal optimal setting, but you can control sensitivity with vadThreshold.

[never_is]

thanks for your answer, @Halle

[Author]

Posts