Skip to content

Latest commit

 

History

History
78 lines (54 loc) · 2.38 KB

intro.md

File metadata and controls

78 lines (54 loc) · 2.38 KB
Raw

Lec1 Introduction to BCI

Definition

  • The goal of BCI was to give significantly paralyzed people another way to communicate, a way that does not depend on muscle control.
  • A system which takes a biosignal measured from a person and predicts (in real time / on a single-trial basis) some abstract aspect of the person’s cognitive state.
    • f: biosignal -> cognitive state: in real time or on a sing-trail basis

Types

  • Active BCI: derive output controlled by user consciously but independently from external events
  • Reactive BCI: reaction to external stimulation
  • Passive BCI: collecting arbitrary brain activities, can be parallelizing

Input: bio-signals

Brain signals

  • EEG: electrical field, w/ w/o gel
  • fNIRS: blood oxygen level, low resolution in time and space
  • ECoG, Microarrays, Neurochips: invasive brain signals
  • MEG, fMRI: magnetic field

Non-brain signals

  • Motion capture
  • eye capture
  • EMG, ECG, EOG
  • system/ application state: stimulus, current vehicle speed, ...
  • environment signals: temperature, ...

Output: cognitive state

  • tonic state: relaxation degree, cognitive load, ...
  • phasic state: switching attention, type of imagined movement, ...
  • event-relate state: surprised or not, committed error, event noticed or not, ...

Applications

  • communication and control for severely disabled
  • operator monitoring: braking intent while driving
  • forensics: lie detection, brain fingerprinting, trust assessment
  • entertainment: mood assessment,
  • Health: sleep state, neurorehabilitation
  • Social: neuro-wear

Challenge

  • interdisciplinary: signal processing, ML, neuroscience, cognitive science
  • variability: person/task specific
    • different folding of cortex between people
    • diff relevant functional map
    • diff sensor loc
    • variable brain dynamics
  • signal-to-noise ratio is high
  • specific measures are hard to obtain with coarse-grained sensing
  • underlying phenomena of brain
  • EEG is complicate, all sensors record the same signal(superposition of all brain activity), can be handled(linear)

Method

  • sophisticated signal processing
  • statistical approach
  • calibration before using

Tools

  • BioSig: Matlab, old, offline, not easy to use

  • BCI2000: C++, online, good doc, lack of signal processing

  • OpenViBE: C++, visual programming, hard to extend

  • g.BSanalyze: Matlab, commercial

  • BCILAB: Matlab

    tool