I can’t wake up 'Rhasspy' vocally

Hello.

I have an issue :
I can’t wake up ‘Rhasspy’ vocally.”

Hardware: 1 rpi30(local ip ending in 30) 3B+ under buster with sound card respeaker 2 mics and small speaker connected to the respeaker + Jeedom v4 + Rhasspy version 2.5.5 fr in master without satellite, installed with Docker and without Snips assistant.

  • 1 rpi41 3B+ under stretch + Jeedom v3 equipped with rfxcom, zwave zigate and all home automation. The 2 rpi are connected by Jeelink (jeedom).

I tested 2 intents [AllumerEteindre] living room lamp and [OuvrirFermer] laundry shutter. My voice commands work well via Node-Red, but I have to click on the yellow button «Wake Up» to wake up Rhasspy. Please note that the “!” service is green (service activated).

I followed the procedure of the Snowboy documentation to recover 3 models (bonjour, jarvis, Gladys in pmdl) on my laptop, then transferred these files by sftp to rpi30 in/home/pi/.config/rhasspy/profiles/fr/snowboy/. None of these files work.
I «played» with the sensitivity in the rhasspy web interface which updates well in the file ‘profile.json’ , but ‘Wake up’ does not work.

I’m attaching Rhasspy’s log:

[DEBUG:2020-10-20 14:07:57,173] rhasspyserver_hermes: <- HotwordDetected(model_id=‘bonjour’, model_version=’’, model_type=‘personal’, current_sensitivity=0.5, site_id=‘Master’, session_id=None, send_audio_captured=None, lang=None)

My rhasspy settings :
MQTT : internal host local p:1883
Audio recording : arecord
Wake word : Snowboy (bonjour.pmdl).
Speech to text : Kaldi
Intent Recognition : Fsticuffs
Text to speech : PicoTTS
Audio Playing : aplay
Dialogue Management : Hermes MQTT
Intent Handling : Remote HTTP

Can you help me please ?

Set Dialogue Management to Rhasspy and then please post a bit more log.
The logging you provided shows the hotword is detected

I set up with rhasspy, I hear the 3 beeps wav hi lo and error .

[DEBUG:2020-10-20 18:21:42,327] rhasspydialogue_hermes: Publishing 88 bytes(s) to hermes/asr/stopListening
[DEBUG:2020-10-20 18:21:42,330] rhasspydialogue_hermes: -> DialogueSessionEnded(termination=DialogueSessionTermination(reason=<DialogueSessionTerminationReason.INTENT_NOT_RECOGNIZED: ‘intentNotRecognized’>), session_id=‘Master-bonjour-78b54c53-1d7d-47d4-b47b-779f2513f397’, site_id=‘Master’, custom_data=‘bonjour’)
[DEBUG:2020-10-20 18:21:42,330] rhasspydialogue_hermes: Publishing 163 bytes(s) to hermes/dialogueManager/sessionEnded
[DEBUG:2020-10-20 18:21:42,333] rhasspydialogue_hermes: -> HotwordToggleOn(site_id=‘Master’, reason=<HotwordToggleReason.DIALOGUE_SESSION: ‘dialogueSession’>)
[DEBUG:2020-10-20 18:21:42,333] rhasspydialogue_hermes: Publishing 49 bytes(s) to hermes/hotword/toggleOn
[DEBUG:2020-10-20 18:21:42,330] rhasspywake_snowboy_hermes: <- HotwordToggleOn(site_id=‘Master’, reason=<HotwordToggleReason.PLAY_AUDIO: ‘playAudio’>)
[DEBUG:2020-10-20 18:21:42,331] rhasspywake_snowboy_hermes: Enabled
[DEBUG:2020-10-20 18:21:42,337] rhasspyasr_kaldi_hermes: <- AsrToggleOn(site_id=‘Master’, reason=<AsrToggleReason.PLAY_AUDIO: ‘playAudio’>)
[DEBUG:2020-10-20 18:21:42,338] rhasspyasr_kaldi_hermes: Enabled
[DEBUG:2020-10-20 18:21:42,340] rhasspywake_snowboy_hermes: <- HotwordToggleOn(site_id=‘Master’, reason=<HotwordToggleReason.DIALOGUE_SESSION: ‘dialogueSession’>)
[DEBUG:2020-10-20 18:21:42,341] rhasspywake_snowboy_hermes: Enabled
[DEBUG:2020-10-20 18:21:42,342] rhasspyasr_kaldi_hermes: <- AsrStopListening(site_id=‘Master’, session_id=‘Master-bonjour-78b54c53-1d7d-47d4-b47b-779f2513f397’)
[DEBUG:2020-10-20 18:21:42,343] rhasspyasr_kaldi_hermes: Stopping listening (session_id=Master-bonjour-78b54c53-1d7d-47d4-b47b-779f2513f397)
[DEBUG:2020-10-20 18:21:42,440] rhasspyasr_kaldi_hermes: Receiving audio

voice commands do not work.
Thank you for your reply.

This is why the command do not work, your intent is not recognized.

Can you post your sentences? Maybe try another setting for Intent Recognition

[AllumerLumiere]
Allume la lumière (du | de | de la | des | dans l| dans la | dans le ){piece} [en ($color){couleur}]
Allume le ($piece)
Met de la lumière
Allume la lumière du salon
Met de la lumière dans le salon
Allume la lumière de la cuisine
Met de la lumière dans la cuisine

sometimes it works but not often.
So we have to set the dialogue management with rhasspy ?

Yes, the dialogue manager is the sort-of-controller. Without a dialogue manager, there is no interaction between the services.

So you say sometimes it works? If it sometimes works and often not, I suggest using another setting for Intent Recognition and see if you get better results

Hello.
The Intents are well recognized with ‘hermes mqqt’ in the ‘management dialog’ with 'Fsticuffs’by using the yellow button ‘Wake Up’
As soon as i set the 'management dialog’with ‘Rhasspy’ it does not work and the 3 beep wav hurt the ears.
I’ve tried with ‘fuzzywuzzi’ and ‘Rasa NLU’ without results.
Just so you know, during the evening my shutter has risen without having ordered it vocally, propibly due to the noise of the tv, i poweroff the rpi for the night.

When you say your voice command (with dialog managment set to rhasspy) do you leave a short pause between the wakeword and the voice command ?

It would be usefull to take a look at the intent message you get when no intent is recognized so you can see what text it actually understood.

Hello to all.
The Intents are well recognized with ‘hermes mqqt’ in the ‘management dialog’ with 'Fsticuffs’by using the yellow button ‘Wake Up’
As soon as i set the 'management dialog’with ‘Rhasspy’ it does not work and the 3 beep wav hurt the ears.
I’ve tried with ‘fuzzywuzzi’ and ‘Rasa NLU’ without results.
Just so you know, during the evening my shutter has risen without having ordered it vocally, propibly due to the noise of the tv, i poweroff the rpi for the night.

I understand that this is no longer a technical problem but I have to find the right setting.
In the past, when I used snips we heard a beep after wakeword.
Can we trigger a low wav beep after the wakeword is pronounced in Rhasspy ?. I haven’t read anything about it, indicate me please, where i can read a tutorial.
thank you again for your answers.

Rhasspy has a sound after the wakeword is recognized by default and it only records after it has played (make sure it is a really short sound). I also use the LEDs on my respeaker 2 mic hat as an indication for wakeword detected. To activate them i use node-red and the websocket listening to /api/events/wake which will then trigger the LEDs.

You can change the wav file played for wakeup, end of command and error.
Using the webinterface you can change the path for the file that should be played for each of them under the settings tab.

1 Like

Thank you very much moquart :grinning:

Good evening.

I just saw your message. I fix the dialogue management with rhasspy and I changed the wake-word by blueberry in porcupine that seems more stable than snowboy.
Now I have a sound problem with a satellite rhasspy. I post a new topic.

thank you very much.