Voice Activity Detection

4 posts / 0 new
Last post
alawre68
Offline
Last seen: 3 weeks 1 day ago
Joined: 09/17/2018 - 10:00am
Voice Activity Detection

I am trying to get some kind of voice activity detection running. I see that the Sensory THF modules have a speech active output. Is there any stand-alone VAD, or is this the only module that currently does this? Is it possible to get 'empty' model, search, and grammar files so that I can test the speechActive output?

PaulB
Offline
Last seen: 18 hours 53 min ago
Joined: 10/04/2016 - 11:43pm

The Sensory THF module requires a trigger model in order to initialize and run.  You won't be able to use it as a VAD-only module.

There isn't a dedicated VAD module in Audio Weaver.  The "Quiescent Sound Detector" can be tuned and work like a basic VAD.

alawre68
Offline
Last seen: 3 weeks 1 day ago
Joined: 09/17/2018 - 10:00am

Do you have any advice on how to tune the QSD to work more like a VAD? Currently I'm using a NoiseGateCore and SlewRateLimiter, on a signal with noise reduction and band-passing of the vocal range. Do you think the QSD could significantly out-perform that setup?

Tunno
Offline
Last seen: 6 months 3 weeks ago
Joined: 01/16/2017 - 2:26pm

QSD is a simple sound detector with band-passing voice frequency range. I think QSD performance should be about the same as your implementation. Sensory voice activity detector performance is about the same as well.