Read the statement by Michael Teeuw here.
MMM-voice
-
@gismo2006 here are some tips to improve the experience:
- it’s necessary to have a proper power supply, in order to decrease the response time, otherwise the cpu will be underclocked and this is contra productive for an cpu intensive task (for rpi 3 it’s recommended to use at least 2.5A for normal usage, I’m using 3A)
- look into the MMM-voice.dic file how you need to spell the words correctly, the language model is based for native speakers and dialects can lead to false positives.
- a cheap microphone can lead to false positive as well, by sending an unclean audio stream.
- speaking louder gave me better results.
- pocketsphinx also tries to match background noise to words, so short words like on, off, one, etc. are getting recognized without saying anything
- so those false positives get spread within your commands and will lead to unwanted actions e.g. you say
MAGIC MIRROR VOICE
to enter the voice mode it propably recognizesMAGIC ON MIRROR VOICE
and then the mirror wont start because the keyword must be recognized together without any other word between or it recognizesMAGIC MIRROR OF VOICE
and then it will not switch to mode voice because the mode has to be right after the keyword or sentence - the error log gets created from pocketsphinx, so those numbers are not from my side
- in order to test what the mirror detects, you can use the config option
debug: true
but don’t use this in production it will generate a huge logfile over time
-
@strawberry-3-141 Thank you for your tipps. I didn’t mean anything angy. I know it is a hard work to code something like that. And I really happy and thankfull that there a coder like you which share it whit the community. And helping if there are any questions. :-)
I will test aggain at he few next day, and im sure there are somme new questions.
Greets gismo
-
I’ve the next problem: On the Mirror shows the message ‘Error’ , ‘No Moduls identified’.
in the Error Log: Error: spawn pocketsphinx_continuous ENOENT-2help me pls ;(
his problem was that somehow his
~/.bashrc
file got overwritten and then he missed the path to pocketsphinx files -
Hmm… getting errors when trying to install the dependencies … can you advise is this known?
-
@wjdw87 Looks like you’er using a Mac. Not sure about Mac has a Package Manager as apt-get, the installer was Designed for raspbian
-
ah… yes using mac… (will move over to windows based later, but just getting all the customisation out of the way on here and learning how it all works before doing additional functions etc)… anyway… i take it there is no ‘easy/guided way to install on here?’
Thanks @strawberry-3-141
-
@wjdw87 you can read your way through the installer and install the dependencies manually, only the first step is based on apt-get, maybe the rest is working on a Mac as well, then you can comment out step one and rerun the installer
-
Hello,
I’m trying the MMM-voice but no luck for now.
I got “Error” on the Magic mirror and i have the following error in the debug.log :
INFO: continuous.c(295): Specify ‘-infile <file.wav>’ to recognize from file or ‘-inmic yes’ to recognize from microphone.
ERROR: “cmd_ln.c”, line 641: Duplicate argument name in arguments: -adcdev
ERROR: “cmd_ln.c”, line 704: Failed to parse arguments list
INFO: continuous.c(295): Specify ‘-infile <file.wav>’ to recognize from file or ‘-inmic yes’ to recognize from microphone.And got “12712712712712712712712711111111111111111111” in error.log
I already tried couple of things :
- reinstall the module by deleting the node_modules folder and rerun the dependencies.sh script.
- upgrade everything with apt-get upgrade
- Configure the usb microphone by creating .asound.rc file
I’m using a standart usb microphone (http://www.banggood.com/fr/USB-Microphone-For-Raspberry-Pi-p-993726.html)
I’ve found couple of differents versions how to setup an usb microphone on a RPI3 but no luck either to have a clear sound when i’m recording wav file to test it with following command :
- arecord -D plughw:1,0 test_mic.wav
Is there a excellent microphone for RPI3 ?
Really looking forward to use this module.
Thanks for your help
-
With @strawberry-3-141 help, thanks again by the way, it’s working now :)
If you got the same issue, here the solution :
- delete the node_modules folder and then rerun the installer.
- remove the words.json file on root module folder and restart the magic mirror.
(It’s going to be regenerated when the module starts)
That should be it.
-
Hey guys,
i’ve installed the plugin but getting this error:
pocketsphinx_continuous: error while loading shared libraries: libpocketsphinx.so.3: cannot open shared object file: No such file or directory
Any suggestions?