give multiple stims to the Stim Based Epoching box

[KJeffrey]

Thomas,

I went back to trying the SVM classification, and got it working better this time. If I can use it, I will not have to worry about writing any Python scripts. It runs very quickly now, probably because Sebastian put an end of file mark in the EEG file. The attached scenario [smok_clas_svm], attached, works with a short recording we did Sep 1, but not with the EEGs we did in June. Can you tell me how to add an end of file to the end of these prior recordings, or create one using the KB stimulator?

Sebastian found a thread about scilit-learn SVM, which we can now try as an alternative, but I noticed this in one of your responses:

>Since OpenViBE 3.3.0, it is possible to give multiple stimulations to the Stimulation Based Epoching box,
>therefore you could epoch for the Right and Left through the same branch

I tried to modify the working scenario to accomplish this (see attached smok_clas_svm_1path], but I know I am missing something to pass the second class stimulation to the classifier. Am I on the right track?

Thanks,
-Jeff

smok_clas_svm_works.png (30.87 KiB) Viewed 8647 times

smok_clas_svm_1path attempt.png (35.67 KiB) Viewed 8647 times

[KJeffrey]

This did not attach to the previous message: smok_clas_svm
-Jeff

[Thomas]

Hi Jeff,

I'm attaching an update of your scenario with one branch.

I've basically added a stream switch, configured for feature vectors, that will also receive the stimulation for right or left, and output the incoming stream on output 1 or 2 accordingly.

Let us know how this goes!

Cheers,
Thomas

[KJeffrey]

Thomas,

I tried it and after it read the 3 minute file in I got this:

[ INF ] Initialization took 1958 ms
[ ERROR ] At time 183.984 sec <Box algorithm::(0x00006520, 0x00005569) aka SVM classifier> {Error description} : {Received fewer examples (0) than specified partition count (10)}, {Error type} : {Kernel::ErrorType::BadInput (code 51)}, {Error location} : {E:\ovm-m\sdk\plugins\processing\classification\src\box-algorithms\CBoxAlgorithmClassifierTrainer.cpp::297}
[ ERROR ] {Error description} : {Box algorithm <SVM classifier> processInput function failed}, {Error type} : {ErrorType::Internal (code 2)}, {Error location} : {E:\ovm-m\sdk\kernel\src\kernel\player\ovkCSimulatedBox.cpp::168}
[ ERROR ] {Error description} : {Process failed for box with id (0x00006520, 0x00005569)}, {Error type} : {Kernel::ErrorType::Internal (code 2)}, {Error location} : {E:\ovm-m\sdk\kernel\src\kernel\player\ovkCScheduler.cpp::578}
[ ERROR ] Scheduler loop failed.

-Jeff

[KJeffrey]

Thomas,

Ignore the title of this topic, I am concerned about the 2nd question I asked, how to use a single branch to use for both stim types. You did not reply to my last error report, and I went in other directions and did not follow up. I think if we can get this working it might help solve the other problem with feature selection.

-Jeff