I’m interested in this subject, the use of 8 micarrays on jarvis. I have recently bought a matrix creator and I want to use it with jarvis. I have well installed the matrix OS, no problem so far. I have tested the arecord command in all 8 MIC channels, for example, arecord --device=mic_channel1 -r 16000 -c 1 -f S16_LE test.wav it works fine, I can record myself even when there are some music in the room and when I play the sound is clear.
However, I have not yet succeeded to trigger a hotword to wake-up snowboy.
I was wondering, if @LengZai from this thread https://github.com/matrix-io/matrix-creator-malos/issues/23#issuecomment-259837301 or anyone else got it worked. Any help or suggestions will be much appreciated.
I think its a general problem for most. I really hoped more would happen in these forums, but they pretty much seem dead. I know I cannot make it happen on my own. Its some nice things the creators have made for it, but hard to find something that can work together with their tools that is voice activated and not much help to find here.
Hopefully more will happen when the Matrix Voice comes out, and it is backwards compatible.
That’s what it seems, people are so silent in this forum. I don’t know if it has worked for them. In my case I’m using Snowboy as stt engine but I fail to have a natural on board phrase detection (it works only when I record manually)…it’s like the mic array doesn’t hear (capture) my voice.
They got hands free working with Alexa I see:
The last tutorial I found needed you to first move your hand above the Matrix Creator and then talk to Alexa. But I also want a keyword based offline mode for stuff other than Alexa, that would be extremely useful. Hopefully when I have more time I will try again.
Thanks for sharing, though me too I saw it already but I have not yet tested as for the moment, I prefer the localized stt engines like snowboy and not the AWS. But I’ll try it one day.
I’m glad that I’ve succeeded to use the Matrix Creator’s mic arrays with snowboy. I had to use arecord piping from a PR on github to solve pyaudio/portaudio problems. Snowboy can detect my hotword. Now I can continue with my project.
Awesome That is what I have been hoping to do as well. I might have time this weekend to give it a try. I would like to have Alexa also, but it should work as an app you can run using a keyword with Snowboy, if that is the one I get to work. Hopefully one of us can make a tutorial on how to set it up later, to get this community going a bit more.
That’s awesome.All the best on your weekend project.
I got the hands free version of Alexa working. It should allow to use Snowboy for the wakeword, but for some reason with the Matrix they only recommend Sensory. I guess I should try to find out if it would be easier just to use Sensory.
Otherwise, do you have a link for that arecord github rep?
Thanks. Might not get working much more on it till next weekend though.
Hi all, I’m a tad late to the game, but what quality of audio are you getting when you use the arecord pipe? I bought the matrix voice primarily for the far field mic array. Does all the signal processing happen automatically when you use mic_channel1? Right now my performance is no better than a standard mic…
My experience is that the full array of mics using pipes was brilliant; issue now is that if you want to use it, many other things don’t work. But soon they say, soon…
@OPK All i really need is the audio, is there somewhere that describes using arecord to get the full array of mics as a single processed stream?