16.11.2012 Views

Brain–Computer Interfaces - Index of

Brain–Computer Interfaces - Index of

Brain–Computer Interfaces - Index of

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

294 C. Guger and G. Edlinger<br />

Fig. 10 Typical experimental<br />

workflow for EEG data<br />

acquisition using spatial<br />

pattern calculations (CSP)<br />

without feedback (FB) and<br />

weight vector (WV)<br />

calculation for the sessions<br />

with feedback<br />

CSP1<br />

CSP2<br />

CSP3<br />

Session 1 without FB<br />

Data used to set up CSP1<br />

and WV1<br />

1st day<br />

Session 2 with FB<br />

2nd day<br />

Session 3 without FB<br />

Data used to set up WV2<br />

2nd day<br />

Session 4 with FB<br />

Data used to set up CSP2<br />

and WV3<br />

2nd day<br />

Session 5 with FB<br />

Data used to set up CSP3<br />

and WV4<br />

3rd day<br />

Session 6 with FB<br />

3rd day<br />

WV1<br />

WV2<br />

WV3<br />

WV4<br />

(event-related desynchronization and synchronization) can be calculated to identify<br />

ERD/ERS components during the imagination for further real-time experiments<br />

with feedback [19].<br />

Based on this knowledge e.g. the band power is computed in the alpha and beta<br />

bands <strong>of</strong> the EEG data. This is done by first band pass filtering the EEG data, then<br />

squaring each sample and averaging over consecutive samples for data smoothing.<br />

This results in a band power estimation in the alpha and in the beta range for each<br />

channel. These signal features are then sent to a classifier that discriminates left from<br />

right movement imagination. As a result, a subject specific weight vector (WV) as<br />

illustrated in Fig. 10 is computed. This weight vector can be used in the next session<br />

to classify in real-time the EEG patterns and to give feedback to the subject as shown<br />

in Fig. 9.<br />

The Simulink model for the real-time analysis <strong>of</strong> the EEG patterns is shown in<br />

Fig. 11. Here “g.USBamp” represents the device driver reading data into Simulink.<br />

Then the data is converted to “double” precision format and connected to a “Scope”<br />

for raw data visualization and to a “To File” block to store the data in MATLAB<br />

format. Each EEG channel is further connected to 2 “Bandpower” blocks to calculate<br />

the power in the alpha and beta frequency range (both ranges were identified<br />

with the ERD/ERS and spectral analysis). The outputs <strong>of</strong> the band power calculation<br />

are connected to the “BCI System”, i.e. the real-time LDA implementation which

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!