Introduction

BCIs

• Brain-computer interfaces (BCIs) have potential to help severely disabled people by translating the intentions of subjects into a number of different commands.
  • Usually tested when environmental and biological
    artifacts are intentionally avoided.

SSVEPs

• Steady state visual evoked potentials (SSVEPs) are particularly attractive due to high signal to noise ratio (SNR) and robustness.
  • In this study, we deliberately introduced different perturbations in order to test the robustness of a steady state visual evoked potential (SSVEP) based BCI.

About

• This is the first study that systematically analyses the effects of these perturbations on the online perform-ance of SSVEP based BCI within and across subjects.

Materials & Methods

Participants

• Recruitied in summer-autumn 2016.
• Most of them were students at Higher School of Economics.
• No payment for the participants.
• 27 healthy subjects between 18 and 41 years of age.
  • 9 males.
  • mean 26 years of age
• All subjects participated in both offline and online tasks.

Experiment Setup

• EEG were recorded in an electrically shielded dark cabin.
• Stimulus paradigms were prepared in Matlab
  software using Psychophysics Toolbox Version 3.
• Participants followed the stimuli presented on a Ultra HD LED Monitor.

Experiment setup

• The main stimuli are composed of four circles placed in different locations with individual flickering frequencies.
  • Up: 5.45 Hz
  • Down: 8.57 Hz
  • Right: 12 Hz
  • Left: 15 Hz

Offline Task

• For each of 25 trials, there is an instruction on the screen informing a subject to focus on the presented circle.
• Subjects then focus on each of the four randomly presented flickering circles indicated by a red oval frame for three seconds with an inter-stimulus interval (ISI) of one second.

Online Task

• Subjects focuses on one of the four flickering white circles for 3 seconds.
• Classification result is presented on the screen with a red color.
• Subjects either confirm or reject the location using keyboard.

Online Task

• The online task was performed under four conditions randomized across subjects:
  • Control
  • Speaking
  • Thinking
  • Listening

Power Spectrum

• Fast Fourier Transform was used with Hanning window to calculate the power spectrum of the preprocessed EEG for a 3s stimuli length.
  • delta (1±3 Hz)
  • theta (4±7 Hz)
  • alpha (8±13 Hz)
  • beta (14±30 Hz)

Classification

• There are 3 classifiers were used to evaluate:
  • Decision tree
  • Naïve Bayes
  • K-Nearest Neighbor

Results

Offline Classification

Power of Spectrum

Online Classification

Conclusions

Conclusions

• SSVEP-based online BCI under different perturbations is robust.
• loud or silent counting resulted only in slight decrease in BCI performance.
  • Indicates that SSVEP-based BCI can be used in parallel during the conversations.
• Some subjects perturbations resulted even in better performance.
  • The different cognitive strategy can be used for improving the accuracy of BCI within individuals.