Mycroft seems slower to respond on the Mark II

[krisgesling]

Describe the bug
Have had it raised by community members that responses to queries sometimes seem very slow on the Mark II where they are not on other instances of Mycroft. It seems that this may be caused by the listener silence detection not working correctly.

By default Mycroft will listen for up to 10 seconds before sending this to an STT service for transcription. However as many utterances are much less than this, if there is a long enough period of silence, Mycroft will consider the utterance finished and send the audio earlier than that 10 seconds.

I'm seeing that sometimes the listener remains open for the whole 10 seconds, even if the user is not speaking. This makes it seem like the processing is taking a very long time but it's actually listening for too long.

I presume this is related to the silence detection and may differ based on the audio input being received.

If that's the case, consideration should be given to making whatever fixes it configurable so it can be defined in mycroft.conf or the Hardware Abstraction Layer and hence differ from device to device.

To Reproduce
Steps to reproduce the behavior:

1. Open the mycroft-cli-client
2. Speak to Mycroft
3. Watch when the mic is activated and stopped.

[ChanceNCounter]

+1 Gez's suspicion that it's related to silence detection, as I've noticed the record phase lasting all 10 seconds a number of times. I hadn't reported this because I assumed it was just me; I have mine on a floor of the house with lots of fans, and it's not every time.

[j1nx]

Isn't the sound of the Mark-2 fan causing this? Perhaps have a look at the CPU temp / LED to see if it happens when the fan is at maximum speed.

[ken-mycroft]

This is an astute observation and actually where I was planning on starting. The mic input level adjustment code is probably having a difficult time with the fan at speeds above 50%. TY for the input :-)

[j1nx]

Luckily easy to initially test. Just start by recording some WAV files with the fan connected and disconnected.

To be honest, at max speed the fan is relativily noisy. Especially if you consider there are only a few centimeter between the fan and the right mic on the board.