Hello Lucy not detecting

[varadwattamwar]

I completed all the steps successfully but after speaking Hello Lucy there is no response from MM. Mic and speaker is set properly.

[sdetweil]

have you tested the mic this way

arecord test.wav
speak something
ctrl-c to stop recording
aplay test.wav

do you hear what you said?

[varadwattamwar]

aplay test.wav

Yes, I heard what i said, but there is some noise disturbance also.

[sdetweil]

then it will be hard for computer to sound match.. u need to get rid of the noise

[varadwattamwar]

then it will be hard for computer to sound match.. u need to get rid of the noise

Ok alright, need to try with a new mic then.
Thanks for the reply.

[varadwattamwar]

then it will be hard for computer to sound match.. u need to get rid of the noise

then it will be hard for computer to sound match.. u need to get rid of the noise

Ok alright, need to try with a new mic then.
Thanks for the reply.

Still the result is same , no response from lucy.
One thing I observed is, I also get an error for Hello-Lucy while running MM.
Error: Failed to fetch git data for Hello-Lucy: Error: block timeout reached
Is this the reason for no response?

[sdetweil]

Error: Failed to fetch git data for Hello-Lucy: Error: block timeout reached
Is this the reason for no response?

the timeout on the updatenotifications module is too short.. fixed in the upcoming release, and in the magicmirror develop branch now

when MagicMirror is running, in a terminal window or ssh session window do

ps -ef | grep arecord | grep -v grep 

it should put out a line like this

????     11505 17306  0 05:11 pts/9    00:00:45 arecord -D default -q -r 16000 -c 1 -t wav 

if it does not, then it means arecord failed to start
in my case it is using the default device...
arecord -l (lower case L, for list)
will show the devices... i don't know how Lucy is configured for mic..

[varadwattamwar]

Error: Failed to fetch git data for Hello-Lucy: Error: block timeout reached
Is this the reason for no response?

the timeout on the updatenotifications module is too short.. fixed in the upcoming release, and in the magicmirror develop branch now

when MagicMirror is running, in a terminal window or ssh session window do

ps -ef | grep arecord | grep -v grep 

it should put out a line like this

????     11505 17306  0 05:11 pts/9    00:00:45 arecord -D default -q -r 16000 -c 1 -t wav 

if it does not, then it means arecord failed to start
in my case it is using the default device...
arecord -l (lower case L, for list)
will show the devices... i don't know how Lucy is configured for mic..

There is only one device mic shown in the list whose card and device number is 1,0.
Still lucy is not detecting users voice so any other solution?
My Rpi version 3B+.

[varadwattamwar]

I tried:
In asoundrc file I changed “hw:1,0” to “plughw:1,0”
But the result is same lucy is not responding.
Please can you give another solution @mykle1

[sdetweil]

after changing .asoundrc did u reboot and also do the arecord test.wav/aplay test.wav cycle?

[varadwattamwar]

arecord test.wav shows a error:
ALSA lib conf.c:1891:(_snd_config_load_with_include) :3:1:Unexpected }
ALSA lib conf.c:3650:(config_file_open) /home/pi/.asoundrc may be old or corrupted: consider to remove or fix it
ALSA lib conf.c:3572:(snd_config_hooks_call) function snd_config_hook_load returned error: Invalid argument
ALSA lib conf.c:4026:(snd_config_update_r) hooks failed, removing configuration
arecord: main:828: audio open error: Invalid argument

So should I change “plughw:1,0” to “hw:1,0”

And what should I do to get lucy working??
Please help!!! @sdetweil

[sdetweil]

i would change it back to how it was.. and reboot..

i do not know.. voice detection is done by an outside module.. pocket_sphinx_continuous
i stopped using Lucy because it didn't recognize my speech 90% of the time.

you could add

debug:true 

to the config:{} section for the Hello-Lucy module in config.js and
then check the window where u start MM, either npm start or pm2 (using the pm2 logs --lines=100 command)
to see what, if anything, was recognized..

if you use npm start, u can redirect the output to a file , with npm start >somefile
and then u can look at somefile live or after stopping MM..

[varadwattamwar]

I identified the issue .
The issue is lucy is not correctly interpreting my voice.
Ex: If I say "sleep" It interprets "eight"
This is the issue
Is there any solution for this ?
Is this the fault of the repository or mic?
But my mic is fine.

i would change it back to how it was.. and reboot..

i do not know.. voice detection is done by an outside module.. pocket_sphinx_continuous
i stopped using Lucy because it didn't recognize my speech 90% of the time.

you could add

debug:true 

to the config:{} section for the Hello-Lucy module in config.js and
then check the window where u start MM, either npm start or pm2 (using the pm2 logs --lines=100 command)
to see what, if anything, was recognized..

if you use npm start, u can redirect the output to a file , with npm start >somefile
and then u can look at somefile live or after stopping MM..

[sdetweil]

according to the cmu sphinx doc, you could try to create a new acoustic model

https://stackoverflow.com/questions/33453691/cmu-sphinx-for-indian-english

which leads to this
https://cmusphinx.github.io/wiki/tutorialadapt/
and here
https://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/

lucy module expects the ldc/lm files to be named Hello-Lucy.dlc and Hello-Lucy.lm
but I think it regenerates based on the word.json file for usage specific recognition
using lmtool (http://www.speech.cs.cmu.edu/tools/lmtool-new.html)

I don't know how to do the adaptation

[varadwattamwar]

word

Thankyou will try this.

[cq0q]

word

Thankyou will try this.

Hi sir, i have the same problem now .. which is when i open my magic mirror it shows up: command something then it shows error .. and i can't say anything ..
Did you solve the problem ? If you did please tell me how