xDawn P 300 online not working

[AlxndrP]

Hello !
I have an Emotive Epoc headset which I try to use with the last version of OpenVibe on Ubuntu. I got the data stream live to OpenVibe with a telent raw server, then I launch the p300-speller-aquisition, then the p300-speller-train-xDawn, then the p300 speller classifier. Everything works fine so far and I get classifications around 80%.
But when I launch the p300-speller-online, I get an error saying that the xDawn Spatial filter is getting 14 inputs but expecting only 3...
I read somewhere else on the forum that this was due to a sampling frequency not being ajusted well during the xDawn training. It seems that the original design was for a sampling frequency of 512 whereas the Epoc Emotive uses 128. The post said to use a factor of 4 for oversampling, but it did not solve the problem.

This is my acquisition servers configuration:
Raw Telnet
...
14 channels
Sampling Freq : 128
...
Limit speed: checked
Big Endian
Header Size 0
Interframe Size 0

with a 4 ms drift tolerance and 4 as oversampling factor.

Any help would be greatly appreciated since I've been trying for several hours without success.

Cheers, Alex

[jlegeny]

Hello AlexandrP,

the scenarios in the P300 chain have all a decimation box with a factor of 4 with exception of the first acquisition scenario. You can remove these decimation boxes instead of adding an oversampling factor.

Please note that the current implementation of the P300 processing has a bug that makes it not always work with the training scenario. For better results please use the online scenario for training as well as online use. You can either disconnect the selection stimulations on the P300 visualisations, or ignore them during the "training" run.

Cheers
Jozef