#Whole Home In-Ceiling Microphones For Voice Control

1 messages · Page 1 of 1 (latest)

pale kiln
#

I am building a large custom home and pulling 30,000+ linear feet of low voltage wire.

I want to be anywhere in my home and control every aspect of my HA smart home with my voice (without being dependent on Internet access/Google/Amazon/Apple/etc.).

Could someone recommend what microphone system I could install throughout my home (ideally in a ceiling) to achieve this goal?

BTW - I've seen some interesting YT videos using Wyoming Satellites and various Pi Zero 2 W + respeaker mics but they are very much in alpha phase.

woven shuttle
#

You may still need to wait. You can take a look at these

royal trench
#

Wait for PE and other user-grade satellites. All things that we have currently are hacky and very limited.

sacred mango
pale kiln
#

I get that Nabu Casa and others are working on PCB dev kits (e.g. esp32 + mic array of sorts running Wyoming protocol). However, wouldn’t it be possible to use dozens of condenser mics (even USB and use a POE adapter) installed throughout my home running into a central rack where an audio mixer would be placed and then feed the audio source to HA (or even a Satellite) listening for the “wake word”?

Granted, only one person in the house could issue a wake word command (which is acceptable) but that way I could install a bunch of mics in my ceilings (and use like HTD’s Alexa kit with a speaker cover) for capturing audio throughout my house? Thoughts?

BTW - I’m okay not knowing exactly where the voice command came from but okay with issuing a command that contains the specific item in a specific room (e.g. “hey Jarvis, turn on master tv and watch XYZ”)

royal trench
# pale kiln I get that Nabu Casa and others are working on PCB dev kits (e.g. esp32 + mic ar...

It's not the question to ask here. It's general question to have multiple microphones with single processor unit. I don't think it's feasible: basically that single processor would be managing dozens of streams simultaneously, searching for the wake word. Then t would have to mute all streams but one to process speech (and again, it's just single mic, which is prone to background noises, unlike arrays that are beaming).

On the other hand, having single processor per house isn't convenient. You might think you're ok with TV, but add "main bedroom right drawer lamp" to your requests and it will be annoying.

Want to do it - try to build it. As of me, it's too hard and miserable. 🙂

true garden
#

Well, in theory the way HA Voice / Wyoming works would allow exactly that. If you have openWakeWord (or even microwave word) running as a HA addon, you can more or less just throw audio streams at it. Look at for example Wyoming satellite on how that can be done

royal trench
true garden
#

Depends on the server 😉

pale kiln
#

On a different thread I was informed that if multiple satellites hear a wake word and send the stream to the HA server, that the HA server would only allow the first satellite (who sent the request first) to send the audio and receive a response from the server. I haven’t yet built multiple satellites to test this yet. If that’s true then only one person in house would be able to issue voice commands…

true garden
#

If two people issue voice commands at exactly the same time, yes. But how often does it happen that people speak the 2 second long wake word at the same time?

pale kiln
#

Well if I use multiple Wyoming satellites throughout my house and two of them hear my wake word at the same time, only one would respond (or so I’m told). The other option I’m exploring is installing regular microphones throughout my house and into a traditional audio mixer and run them all into an audio mixer which would pipe the audio out to HA (or satellite) but I’d need to know from the audio mixer which mic was “the loudest” so I could send the response to the correct in-ceiling speaker zone.

true garden
#

I don't fully understand why you want to separate wyoming satellites to respond to one persons wakeword? That would start two separate pipelines and double the desired action.

royal trench
pale kiln
# true garden I don't fully understand why you want to separate wyoming satellites to respond ...

In order to ensure I have full voice coverage within my home, I want to be sure I have adequate listening devices throughout my home. Thus, my living room area is about 40’ wide and 30’ deep so I know I’d need at least two devices in that area for proper voice coverage but I suspect more than 1 satellite will hear the same command. Whereas if I just use microphones throughout house and connect all of them to au audio mixer than there would only be a single stream of audio (but I get formatBCE point)

royal trench
#

Also with microphones all over, you have to think on analogue signal distortion and level.

Another thing: if your server hears wake word and starts listening, it will still listen all mics, right? Because it can't distinguish exact mic that sent wake word. And if someone in other room will be speaking (not necessarily to the HA), it will interfere with your command, rendering it useless.

royal trench
deft gate
#

Not in celling but my light switch in all rooms

royal trench
#

The question is not if one can do it - but if it's viable solution. 🙂
Does it work reliably with normal speech, across the room, with TV speaking?