silent recording from microphones

[andrewpbstout]

Hi folks,

I'm trying to test Snowboy (http://docs.kitt.ai/snowboy/#running-on-pi) on QT, and I'm failing to get a test recording with 'rec test.wav'. I don't get an error, but when I play back the recording it's silent.

I've tried adding a ~/.asoundrc as they suggest, to no effect:
pcm.!default { type asym playback.pcm { type plug slave.pcm "hw:0,0" } capture.pcm { type plug slave.pcm "hw:1,0" } }

Any hints about how to configure to capture sound from the microphones?

[codionysus]

You can try with this pcm.!default { type asym playback.pcm "plughw:0" capture.pcm "plughw:1" } ctl.!default { type hw card 1 } and if your command doesn't work try this one arecord -D plughw:1,0 --duration=5 test.wav

[andrewpbstout]

Thanks @dkcodeguy ! Your asoundrc doesn't seem to have had any effect, but your arecord command worked.

It turns out that sox needs some environment variables to know the correct source and drivers: setting export AUDIODEV=hw:1,0 and export AUDIODRIVER=alsa allowed me to get the same results with rec as I got with arecord.

But both are VERY quiet. I remember reading something suggesting that was common on RPis; I'll see if I can adjust the gain. alsamixer says that the capture device doesn't have any capture controls...what's The Right Way to adjust the gain? Or is that not actually my problem?

[apaikan]

by VERY quite you mean completely silent (zero data) or very low gain data?
So far i know, for hot world detection or voice recognition, the audio gain is not important. in any case the audio algorithm normalizes the audio signal.
Anyway, for audabilty purpose, you can increase 'AGCGAIN' Respeaker mic array. yoh may also need set 'AGCONOFF'. to do that you need to copy/download tunning.py to RPI to set the Mic parameters. please check the documentation here:http://wiki.seeedstudio.com/ReSpeaker_Mic_Array_v2.0/

[andrewpbstout]

low gain--quiet, not completely silent. I was able to adjust the gain of the recording on the command line so I could hear for test purposes, and adjust the gain via PyAudio when testing Snowboy. Good to have the link to the Respeaker documentation, though!

Good news! After some hacking to get PyAudio to listen to the correct audio device, I got Snowboy's demo working on the robot. That's my plan for a hotword solution for now, but I'd love to know what other people are using.