ivopascal / emg_classifier Goto Github PK

We found that biosemi LSL https://github.com/labstreaminglayer/App-BioSemi
gives data streamed in microvolts, whereas the recording from OpenVibe is in volts.

This means the stream has values 10^6 times higher than expected.
Dividing by 10^6 solved this problem.

We also found that:

• It's needlessly difficult to repeat commands in cmd. Can we set the data file, stream name and ignored channels in another way?
• For some reason we need to reject 33 channels instead of 32. I do not know why. Can we encode this as well?

• Update readme to reflect:

  • Installing windows on Lab computer
  • Installing Pipenv
  • Adding Pipenv to Path if on Windows

• Loosen expecations on Python version

• Have a more elegant way of dealing with Mock channels

• Fix bug to accept ignored channels as str

• Add instructions for running in lab

  • Find openvibe & acquisition server
  • Open acquisition protocol Ivo.xdf (takes 2 minutes instead of 7)
  • Modify stream writer output
  • Run recording
  • Disconnect OpenVibe from EEG
  • Run biosemi.exe
  • Set subset (32)
  • Run Labrecorder to identify CHANNEL_CODE
  • Call program with python filename [CHANNEL_CODE] 32

• Print accuracy before assertion

• Use Createinfo to create fake info

  • ignored_channels + 8
  • sfreq