VAD detects humming as speech

[CuriousMonkey7]

While using the llm_voice_chat demo, I hummed randomly, and the VAD model detected it as speech. However, the STT returned empty text, resulting in this:

I tried setting vad_options.threshold to 0.7 but didn't see any changes. This seems more like a VAD model issue than a FastRC problem. Just mentioning it here in case some parameter can be tuned to resolve it

[freddyaboulton]

Thanks @CuriousMonkey7 - that's interesting. I'll look into it. I think setting a higher threshold should have fixed it though.