Hey there,
I am having trouble with the voice / command recognition with the recently purchased “Home Assistant Voice”.
I tried multiple time with this (“Schalte Bürolicht Schalter an”)
and it only worked after adding an alias and I had to unplug the device / reboot it.
Also I am really having issues with the voice being recognized at all. When I am 4 meters away, the device does not turn on with “Hey Jarvis”. I have to stand 2 meters infront of it, which basically beats the purpose of having a voice control.
Will there be a voice sensitivity bar in the future?
Yeah, it’s just a problem with the XMOS microphone setup in there. All the other devices using it have the same issue (like the ReSpeaker). Buzzwords aside, it’s just a cheap way to avoid using a proper matrix mic setup like conference speakers, Echo or Nest devices use. The problem is the gain control is bad, so the quieter you are (either talking quietly, or standing farther away), the noisier the input gets when it boosts the volume, and it doesn’t take much for these neural-net based STT engines to just not understand. It’s even worse with microwakeword, as it is a very simple engine. I found switching mine to openwakeword helped a lot, but Whisper still gets the conversions wrong a lot of the time if I’m standing too far away. In the end I switched back to using micro wakeword because it not waking up was a good clue that I was too far for the speech to work, anyway.
To be fair, though, they did make it clear these are preview devices. IMO, before they call them “ready”, they’re going to need to rev the hardware and use a real matrix mic setup.
