PusleAudio causes stt-no-text-recognized. doesn't wait long enough to hear command after wake word detected

[MrSco]

It didn't have this issue when I was using alsa. using the same noise cancellation and sensitivity settings.

Also couldn't get the ducking feature of pulseaudio working; the wake word is just impossible to trigger while music is playing. Is there a way to do this with alsa? Pulseaudio seems like more trouble then its worth.

Here is the debug log from wyoming-satellite.service...

Jan 21 19:20:15 workshopsatellite run[1382]: DEBUG:root:Connecting to wake service: tcp://127.0.0.1:10400
Jan 21 19:20:15 workshopsatellite run[1382]: DEBUG:root:Connecting to event service: tcp://127.0.0.1:10500
Jan 21 19:20:15 workshopsatellite run[1382]: INFO:root:Connected to services
Jan 21 19:20:15 workshopsatellite run[1382]: DEBUG:root:Using webrtc audio enhancements
Jan 21 19:20:15 workshopsatellite run[1382]: DEBUG:root:Connected to mic service
Jan 21 19:20:15 workshopsatellite run[1382]: DEBUG:root:Connected to wake service
Jan 21 19:20:16 workshopsatellite run[1382]: DEBUG:root:Server set: 7268086440527
Jan 21 19:20:16 workshopsatellite run[1382]: INFO:root:Connected to server
Jan 21 19:20:16 workshopsatellite run[1382]: INFO:root:Waiting for wake word
Jan 21 19:20:16 workshopsatellite run[1382]: DEBUG:root:Connected to event service
Jan 21 19:20:34 workshopsatellite run[1382]: DEBUG:root:Streaming audio
Jan 21 19:20:34 workshopsatellite run[1382]: DEBUG:root:Running ['/home/pi/wyoming-enhancements/snapcast/scripts/awake.sh']
Jan 21 19:20:34 workshopsatellite run[1390]: Sun 21 Jan 19:20:34 EST 2024 [awake.sh] ...Starting awake.sh script
Jan 21 19:20:34 workshopsatellite run[1390]: Sun 21 Jan 19:20:34 EST 2024 [awake.sh] ...Starting silence
Jan 21 19:20:34 workshopsatellite run[1382]: DEBUG:root:Muting microphone for 0.8995918367346939 second(s)
Jan 21 19:20:34 workshopsatellite sudo[1394]:     root : PWD=/home/pi/wyoming-satellite ; USER=root ; COMMAND=/usr/bin/pacat --client-name=silence --volume=0 --property=media.role=notification
Jan 21 19:20:34 workshopsatellite sudo[1394]: pam_unix(sudo:session): session opened for user root(uid=0) by (uid=0)
Jan 21 19:20:34 workshopsatellite run[1382]: DEBUG:root:Connected to snd service
Jan 21 19:20:35 workshopsatellite run[1382]: DEBUG:root:Unmuted microphone
Jan 21 19:20:36 workshopsatellite run[1382]: WARNING:root:Event(type='error', data={'text': 'No text recognized', 'code': 'stt-no-text-recognized'}, payload=None)
Jan 21 19:20:36 workshopsatellite run[1382]: INFO:root:Waiting for wake word

I do have some possibly concerning things in the pulseaudio.service log tho...

Jan 21 17:19:09 workshopsatellite pulseaudio[403]: Failed to find a working profile.
Jan 21 17:19:10 workshopsatellite pulseaudio[403]: Failed to load module "module-alsa-card" (argument: "device_id="1" name="platform-3f902000.hdmi" card_name="alsa_card.platform-3f902000.hdmi" namereg_fail=false tsched=yes fixed_latency_range=no ignore_dB=no deferred_volume=yes use_ucm=yes avoid_resampling=no card_properties="module-udev-detect.discovered=1""): initialization failed.
Jan 21 17:19:10 workshopsatellite pulseaudio[403]: Failed to open cookie file '/var/run/pulse/.config/pulse/cookie': No such file or directory
Jan 21 17:19:10 workshopsatellite pulseaudio[403]: Failed to load authentication key '/var/run/pulse/.config/pulse/cookie': No such file or directory
Jan 21 17:19:10 workshopsatellite pulseaudio[403]: Failed to open cookie file '/var/run/pulse/.pulse-cookie': No such file or directory
Jan 21 17:19:10 workshopsatellite pulseaudio[403]: Failed to load authentication key '/var/run/pulse/.pulse-cookie': No such file or directory
Jan 21 17:19:10 workshopsatellite systemd[1]: Started PulseAudio system server.
Jan 21 17:19:32 workshopsatellite pulseaudio[403]: ALSA woke us up to write new data to the device, but there was actually nothing to write.
Jan 21 17:19:32 workshopsatellite pulseaudio[403]: Most likely this is a bug in the ALSA driver 'snd_bcm2835'. Please report this issue to the ALSA developers.
Jan 21 17:19:32 workshopsatellite pulseaudio[403]: We were woken up with POLLOUT set -- however a subsequent snd_pcm_avail() returned 0 or another value < min_avail.

[MrSco]

I believe this is just a side effect of the pulseaudio delay. I start issuing my command a couple seconds later then I normally would after I hear the awake.wav sound because of the delay.

I did notice though, that since the stt has a 'No text recognized' error, the tts-stop-command never fires which keeps the pulseaudio stream ducked. It was only un-ducking audio with successful commands/responses. So I added a error-command option to my service...

...
--error-command '/home/pi/scripts/error.sh' \
...

#!/bin/bash
echo $(date) [error.sh] ...Starting error.sh script
#echo $(date) [error.sh] Playing error.wav...
#paplay --property=media.role=music --raw /home/pi/sounds/error.wav
sudo /home/pi/wyoming-enhancements/snapcast/scripts/done.sh

I tried to have it play a sound on error also, but it would cause the satellite leds to get stuck in streaming/yellow mode until poked from server side or rebooted.

i also made a modification to the 2mic_service.py to blink the leds purple on error for some visual feedback...

...
_PURPLE = (255, 0, 255)
...
elif event.type == "error":
            _LOGGER.debug("Error: %s", event.data.get("text"))
            # Flash
            for _ in range(3):
                self.color(_PURPLE)
                await asyncio.sleep(0.3)
                self.color(_BLACK)
                await asyncio.sleep(0.3)
...

[MrSco]

I ended up switching back to Alsa. Its just so much faster then PulseAudio. its like night and day.

I also enhanced the 2mic_service.py to have some better feedback while waiting for the assist response audio ...

rhasspy/wyoming-satellite#63 (comment)

[MrSco]

Ah thanks. The wake word is never detected while music is playing even at the lowest of volumes so having pulse audio for the audio ducking really doesn't matter as that's all it's used for. No noise cancellation or other settings help that so not worth the trouble. I'm using shairport-sync also which allows airplay in addition to music assistant compatibility. Snapcast is just too slow and pause doesn't work. Also snapcast doesn't seem to have a way to locally play/pause the music on the satellite itself. With shairport-sync I can use a dbus command to play/pause toggle the music by hitting the respeaker button... or through MA. So the alsa/shairport-sync combo just has better functionality vs pulseaudio/snapcast for me since the wake word will never wake a satellite that is playing music. Apparently only Apple, google, and Amazon have the good enough mic arrays to handle that well.