bci2000.org BBS

Dear Gerv,

I have Tried BCI2000 for speller. I set the linear classifier, column 2, 280ms and 580ms. But there is warning ""Channel specification in physical units: 580ms not exactly meet a single channel". Would you please to inform me, what should I do with this warning? Is it OK, if I trial the speller with this condition?

Second question, about P3 temporal filter. Would you please to explain me, what happened if I changed the epoch length at P3 temporal filter section? Is there any effect if I changed this parameter?

Thank you. Please adive.

best regards,
reza darmakusuma

Reza,

The only constraint for the epoch length is that the epoch has to be at least as long as the times you are referring to in the classifier (e.g., 580 ms).

Juergen will get back to you about the other question.

Gerv

Reza,

thank you for asking this question. I have added a note to the P300 tutorial to explain the warning message.

In my answer, I suppose that the warning message refers to "elements" rather than "channels". Please refer back if this is not the case.

The background is the following: BCI2000 allows you to specify physical units wherever possible, and thus tries to simplify configuration, to avoid configuration errors, and to make configuration independent of changes to sampling rate, block size etc.

If you don't like the idea of configuring BCI2000 in terms of physical units, you may always choose to provide 1-based indices. However, changes to sampling rate or block size will then require changes to a number of configuration parameters, and these may be difficult to identify, and cumbersome to calculate.

The "does not exactly refer to" warning message is issued by the LinearClassifier whenever the specification in physical units does not exactly match a single sample/frequency bin/channel in the signal it refers to. E.g., for a sampling rate of 128Hz, individual samples will be 7.8125ms apart, so specifying an offset of 580ms will not exactly refer to a sample offset.

In this case, BCI2000 will pick a sample by rounding downwards, i.e. towards the full sample position located earlier in time, and will issue the warning message.

In the example, the offset would be computed to be 580ms*128Hz=74.24 in terms of a sample's duration, and rounded down to 74. The first sample starts at offset 0, so offset 74 corresponds to sample #75, and its temporal offset is 74/128Hz=578.125ms.

In your case, you may choose one of these solutions:
1) ignore the warning message,
2) write 578.125ms rather than 580ms, or
3) write 75 as a plain number rather than 580ms.

Regards,
Juergen

Thanks alot, for advance advice.


regards,
reza darmakusuma