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?
6:01pm
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.
11:21am
Paul,
In Near end VAD, why are you using Crossover Filter LinkwitzRiley ? Based on which research paper or signal flow diagram,, did you design the Near End VAD for STM32 disc board?
3:36pm
Hi AS. I'm not sure which Audio Weaver signal flow you are referring to. Could you attach the AWD and we'll review it? Thank you.
11:15am
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?
1:18pm
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.