Brain Computer Interface

From Tmplab
Revision as of 19:45, 21 August 2011 by Sameeg (talk | contribs)

/tmp/bci (aka Neurohack)

Project involving hardware and software in 3 different major fields : Science / Computer Interface / Art
Currently 3 hardware and 2 software main directions are followed.

HARDWARE

Commercial ready-to-use BCI-kits

Real brainwaves centered current affordable commercial offers are either too simple (Mindset : ONE dry sensor) or required complex setup (Emotiv: several WET sensors).

NeuroSky Mindset

  • Single EEG dry sensor headset with audio
  • Ability to sample at 512Hz and access the raw signal or filtered FFT transform every second (waves: alpha, beta, delta, ...)
  • BSD license Python SDK that works under linux with serial access mapped to the bluetooth device: https://github.com/groner/pythinkgear
  • Software for OSX and windows also available by headset manufacturer : Neurosky

Emotiv

TODO: describe me!

DIY EEG headset

  • 2 OpenEEG sets were built (2 wet electrodes each)
  • Continuous raw data feed only that must be filtered and analyzed.

DIY Near Infra Red Spectroscopy headset

  • Designing a new acquisition device (efficient, easy to wear, affordable) using Near Infra Red Spectroscopy
  • Prelimanary tests were done with a light fiber and spectrometer
  • Joint developement with Editions Volumique


SOFTWARE

Bainwaves to music, lights, sounds,...

The homebrewed software is made in Rebol and called "Jackson". Several versions have been created for dedicated purposes, like jprocessing or jLIVE, a sequencer style live tool, available from 1 to zillion brains, for art performance, cooperative "construction" (graphics, sound, DMX light control, command,...). The Rebol option was chosen for dead simple portability and to interface seamlessly local and Internet connexions.
Currently supported hardware is the Mindset.

Brainwave sounds

Made in Pure Data

WHAT WE DO

SCIENCE

  • Neurosciences : dedicated software for brain functionnal imaging : MRI, near IR, EEG with machine learning
  • Large sets of brain generated data available in CSV file format. (40 hours)
  • For more information on machine learning see the subproject page: /tmp/bci/machine-learning

Current steps are : designing experimental tests and results statistic analysis to stronger assess hardware reliability.

ART

Neurohack, as Digital arts - music / video live rendering, human performance of the brain activity - has been done in different venues :

BRAIN COMPUTER INTERFACE APPLICATIONS

  • Computer, robots,... command.
  • Help disable people.
  • BYOU : Bring Your Own Use : the geekier, the better !


WHAT NEXT ?

  • Improve OpenEEG solution by signal analysis.
  • Improve control by implementing Machine Learning.
  • Develop artistic collaboration especially in Music composition and wide public use.
  • Continue bootstrapping people that need some experience in our knowledge area.


For more about Jackson : /tmp/bci/jackson.
For more references (bibliography) : see the /tmp/bci/starterkit.
The repository for tmpbci software is tmpbci on github
To join us or for questions, welcomed comments, ideas : tmpbci at gmail dot com
Neurohack meeting is following tmp lab meeting, every thursday 9pm+ at La gaite lyrique, Paris
Last update : August 2011