Wake Word on RPi zero (porcupine) don't work

Hi all,
I changed my satellite to a RPi zero with ReSpeaker 2-MicsPiHat.
The recognition works fine but the detection of the wake word won’t work.
Here the profile:
“intent”: {
“system”: “hermes”
“microphone”: {
“arecord”: {
“device”: “default:CARD=seeed2micvoicec”,
“siteId”: “satellite”
“system”: “arecord”
“mqtt”: {
“enabled”: “true”,
“host”: “192.168.xxx.xxx”,
“port”: “12183”,
“site_id”: “satellite”
“sounds”: {
“aplay”: {
“device”: “default:CARD=seeed2micvoicec”
“system”: “aplay”
“speech_to_text”: {
“system”: “hermes”
“text_to_speech”: {
“satellite_site_ids”: “satellite”,
“system”: “hermes”
“wake”: {
“porcupine”: {
“keyword_path”: “computer_raspberry-pi.ppn”,
“sensitivity”: “0.40”
“satellite_site_ids”: “satellite”,
“system”: “porcupine”

Can somebody help me to solve the problem?


Which OS have you installed ? Is it the Lite or Desktop version of Rasperry Pi OS ?

Have you checked that the microphone and speakers are working ?

Are you using the Rhasspy Docker install, or the Debian install ?

I also have a RasPi Zero with ReSpeaker 2-Mics HAT. The procedure I used is in my tutorial. Not the only way, but what worked for me.

Hi Don,

as a last trail I 'll tested your tutoral. In case it does’nt work there is a RPi0 for sale ;-).

Greetings Roman

I follow your tutorial but unfortunately there some bugs.

arecord -f cd -Dhw:1 test.wav

won’t work output:

arecord: set_params:1407: Unable to install hw params:

Same with the aplay command. The call of the LED test script is not correct, but if I changed to the directory and run the script, the LED lights.

Which deb file do you get? My architecture answered armhf!

dpkg-architecture | grep DEB_BUILD_ARCH=

Should I use the armel.deb?

What happens when you do aplay -l? as do you have a hw:1 and also can it record at cd quality as plughw:1 will do auto conversion.

Usually I test at what Rhasspy normally records with arecord -Dplughw:1 -fS16_LE -r16000 -c1 test.wav the vu is also handy for a test is it -V mono from memory?

So I have done several test and the rhasspy is working with recognition of spoken words. Wake word don’t react in any languages I now?

I switch to Rhasspy Raven wake word detection and it is working, sometimes.

I would check your inputs in terms of input volume Roman as often it can be very low.
Raven is great for collecting KW and as a quick start-up but not that great of a KWS so dunno if its your setup or maybe your volume is low or config bad or its just Raven being Raven.

I investigated deep into the relations between rhasspy and porcupine. The install requirements said the the verson needed is 1.9. I load the wake word from github computer… I ask myself is it possible that the wake word on github is prepared for the latest version of porcupine with access key needed. So that I need a old ppn file or if there is a different between the english and the german wake word file?
Any suggestion?

It was a long hard way to success. But now it is working.
I follow the tutorial by Don and add some details following this site
porcupine update
I had to update the rhasspy twice for add all dependencies. Than add the porcupine accesskey as described.
Now it is running and I am going to add the result to my fhem smart home.
Thanks all

1 Like