use it only for hotword detection

[helloworldkr]

We want to have the only hotword detection separate from avs. We want to use our own speech recognition system. We are finding that difficult.
Please guide us through this

[KillingJacky]

Sorry I can't get your point, could you clarify your expression ?

[sskorol]

@helloworldkr check the project I've created for exactly the same purpose. It streams pre-processed audio chunks from Respeacker Core V2 to the custom ASR engine.

[solidssss]

@helloworldkr check the project I've created for exactly the same purpose. It streams pre-processed audio chunks from Respeacker Core V2 to the custom ASR engine.

Will this stream the audio with AEC and filtering enabled? So I can play music from the respeaker and it will send the filtered audio to the ASR engine?

It would be nice if the respeaker core v2 was compatible with Rhasspy, it works but it does not filter noisy audio from the playback with Rhasspy...Maybe this will somehow be compatible with Rhasspy? Rhasspy also uses websockets

[sskorol]

@solidssss I'm not using Rhasspy. But the library I referenced applies AEC, as well as NS, AGC, WWE, etc. Also note that Respeaker doesn't perform any DSP out of the box. It's possible only via librespeaker library that applies Alango algorithms. So if you just play an audio w/o librespeaker, AEC won't work.