Wake word not working

[ghost]

IMPORTANT: Before you create an issue, please take a look at our Issue Reporting Guide.

Briefly summarize your issue:

I updated the avs to the latest version (I completely deleted the workspace) and since then the wake word doesn't work any more, but t+Enter does perfectly. I accepted the SENSORY-License-Agreements, but I just can't get it to work. "arecord -d 4 -f S16_LE -r 16000 test.wav" just records silence, not even any noises. The volume of my mic (USB: card 1 device 0) is turned to 100<>100 in the alsamixer.

What is the expected behavior?

The wake word should be detected.

What behavior are you observing?

Nothing happens.

Provide the steps to reproduce the issue, if applicable:

Install the AVS using the setup script on a RaspberryPi.

Tell us about your environment:

What version of the AVS Device SDK are you using?

  <1.22.0>

Tell us what hardware you're using:

• [ ] Desktop / Laptop
• [X] Raspberry Pi (2)
• [ ] Other - tell us more:

Tell us about your OS (Type & version):

• [ ] Linux
• [ ] MacOS
• [x] Raspbian Stretch (9)
• [ ] Raspbian Jessy
• [ ] Other - tell us more:

[kclchan]

Hi @JustixDev, the fact that 't+enter' is working for you, it doesn't seem like it's a micophone issue. Could you please attach the DEBUG9 logs for further investigation. Thanks!

[ghost]

Sure, here you go: debug9-log.txt

[kclchan]

Hi @JustixDev, I think this is the problem:

2021-03-24 23:37:34.145 [  1] E SensoryKeywordDetector:initFailed:reason=loadingSensoryModelFailed,error=snsrLoad()\: task\: No such file or directory (open "...lexa-rpi/models/spot-alexa-rpi-31000.snsr")
2021-03-24 23:37:34.146 [  1] E SensoryKeywordDetector:createFailed:reason=initDetectorFailed
2021-03-24 23:37:34.149 [  1] C SampleApplication:Failed to create keyword detector!

Please make sure the file '...lexa-rpi/models/spot-alexa-rpi-31000.snsr' exists. The alexa-rpi path should be under the same path where you specified -DSENSORY_KEY_WORD_DETECTOR_LIB_PATH in your cmake.

[ghost]

Okay I reinstalled the sdk and installed it from source (other guide), but now even t+Enter doesn't work and the logs are getting spammed with readerTimeOut exceptions.

debug9-log-2.txt

Output still works perfectly. This is my ~/.asoundrc file:

pcm.!default {
  type asym
  capture.pcm "mic"
  playback.pcm "speaker"
}
pcm.mic {
  type plug
  slave {
    pcm "hw:1,0"
  }
}
pcm.speaker {
  type plug
  slave {
    pcm "hw:0,0"
  }
}

and this is my sdk-folder/third-party/alexa-rpi/config/asound.conf file:

pcm.dsnooper {
    type dsnoop
    ipc_key 816357492
    ipc_key_add_uid 0
    ipc_perm 0666
    slave {
        pcm "hw:1,0"
        channels 1
    }
}
pcm.dmixer {
    type dmix
    ipc_key 1024
    slave {
        pcm "hw:0,0"
        period_time 0
        period_size 1024
        buffer_size 8192
        rate 44100
    }
}
pcm.!default {
    type asym
    playback.pcm {
        type plug
        slave.pcm "dmixer"
    }
    capture.pcm {
        type plug
        slave.pcm "dsnooper"
    }
}

I use a USB Logitech mic and a JBL Charge 3 as a speaker which is connected to the raspberry pi through the 3.5mm jack.

[ghost]

Now the speaker sounds like Alexa sends the audio in an incorrect rate (see audio test).

audio-test.mp3.zip

[womw]

Hi @JustixDev thanks for posting

Following the steps here: https://developer.amazon.com/en-US/docs/alexa/avs-device-sdk/raspberry-pi.html

Can you update your ~/.asoundrc, your values for pcm.!default are incorrect pcm.!default { type asym playback.pcm { type plug slave.pcm "hw:0,0" } capture.pcm { type plug slave.pcm "hw:1,0" } }

Also can you run sudo apt-get install sox -y && rec test.wav to determine if the microphone works correctly?

Once that is done can you run the SampleApp with the following command cd /home/pi/sdk-folder/sdk-build PA_ALSA_PLUGHW=1 ./SampleApp/src/SampleApp ./Integration/AlexaClientSDKConfig.json ../third-party/alexa-rpi/models

[ghost]

The audio still sounds like the one in the audio-test and now the mic isn't working at all (with the new .asoundrc and the "rec" command.

[kclchan]

This appears to be a configuration issue that is outside the scope of the SDK. I am closing this ticket, please open a new issue if you need more help.