search

openyou / emokit

Open source driver for accessing raw data from the Emotiv EPOC EEG headset
http://www.openyou.org
Other
524 stars 236 forks source link

Emokit+ Artificial neural network? and train them? #129

Closed Joaquimdcp closed 10 years ago

Joaquimdcp commented 10 years ago

We're MovMind, a project to move a robotic arm using emokit. So we've cast a program using neural networks. But the problem is we do not have enough information about the town i on BCI technologies in general. Let us have if you can help you ...

1-We are proving to train neural networks but do not get exactly reproduce the same movements. 2.We haven't got very clear that action would be the most correct 3 values ​​are different town every day. It would be possible to achieve movement patterns to train the neural networks?

Best regards.

ghost commented 10 years ago

This library will give you raw access to the data.

From what I understand you need to record the data while producing input related to the movement.

Maybe do it several times and average the data, if the data falls within a given value range at that point in time say +/-10% of the recorded values then the action matches.

You may have to do some correction in time to get the data to match up properly, like matching a waveform.

What version of the library are you using?

I'm going to be working on training in the emobrain project, although I'm quite busy at the moment with my crm and lawn watering projects... I have to work on the crm before I move on to this.

olorin commented 10 years ago

Yeah, you'll most likely need to apply a Kalman filter or the like before the signal's useful for machine learning purposes, but you should have everything you need.