P300Speller

[BCI2000user]

Hi,
We have tried the P3speller by using the way described which is makng the subject count every time his letter flashes. However, we got accuracy of 0%. When we made the same experiment but we made the subject blinks every time the letter flashes, we got accuracy of 100%. We used Fz as the only channel of recordig after using Mario. Do you have any explanations?

Thank you

[gschalk]

Hi,

You would expect to use more posterior electrodes, such as Cz or Pz, but not Fz. I suggest to first collect a reasonable amount of data, e.g., from 30-50 characters per subject, and then to analyze the data using Mario. Then, compare the responses that you find to known responses in the P300 literature. Only when you in fact confirm that you are looking at a good and reliable P300 response should you go ahead with BCI experiments.

Gerv

[BCI2000user]

what do you think of using one of the artifacts such as blinking or muscle movement as a way to show a response to a stimulus and then use it in an application?

[gschalk]

what do you think of using one of the artifacts such as blinking or muscle movement as a way to show a response to a stimulus and then use it in an application?

Hi,

If you are interested in BCI research, you really have to exclude using muscle activity. If your interest is in producing a muscle-based communication system (which there are many of), then this of course is fine, but you would not be doing BCI research.

The Gerv