BCI Controller for Semi-Autonomous Wheelchair

Working with OpenViBE signal processing scenarios and doing scenario/BCI design
Post Reply
traviskv
Posts: 15
Joined: Thu Aug 02, 2018 12:42 am
Location: USA

Emotiv, OpenViBE, and semi-autonomous wheelchair

Post by traviskv »

Hello, all.

As the title suggests, this post is concerned with implementing a BCI controller for a semi-autonomous wheelchair using an EasyCap/Emotiv headset and OpenViBE to obtain and process signals. The features extracted and classified are sent to LabView (as the semi-autonomous functions and joystick mimic for the chair are written in LabView) via TCP.

My current goal is to use hand motor imagery to turn the chair left/right and foot motor imagery to move forward and stop (or vice versa).

I've gone through the CSP motor imagery tutorials and have successfully created filters and classifications for both hand and foot motor imagery, but is there a way in OpenViBE to combine the CSP filters and classifiers into one script?

I am still new to OpenViBE and BCI controls, so I do not know if this is a feasible idea or if there is another way to accomplish my goals.

Thank you!

traviskv
Posts: 15
Joined: Thu Aug 02, 2018 12:42 am
Location: USA

BCI Controller for Semi-Autonomous Wheelchair

Post by traviskv »

Hello, all.

This post is concerned with a BCI wheelchair control project. I would like to design a BCI controller for a semi-autonomous wheelchair using an Emotiv headset and OpenViBE to obtain and process signals. The features extracted and classified are sent to LabView (as the semi-autonomous functions and joystick mimic for the chair are written in LabView) via TCP.

My current goal is to use hand motor imagery to turn the chair left/right and foot motor imagery to move forward and stop (or vice versa).

I've gone through the CSP motor imagery tutorials and have successfully created filters and classifications for both hand and foot motor imagery, but is there a way in OpenViBE to combine the CSP filters and classifiers into one script?

I've also done a bit of reading on hybrid BCI controllers. Ones using a combination of EEG and EMG signals to issue different commands.

I am still new to OpenViBE and BCI controls, so I do not know if this is a feasible idea or if there is a better way to accomplish my goals. I appreciate any and all information anyone could share with me.

Thank you!

traviskv
Posts: 15
Joined: Thu Aug 02, 2018 12:42 am
Location: USA

Re: BCI Controller for Semi-Autonomous Wheelchair

Post by traviskv »

Perhaps my first post was addressing too many topics! Let me try to reduce my question to a few topics for discussion.

I am designing a BCI controller for a semi-autonomous wheelchair. The current EEG cap I have uses 14 working electrodes and 2 reference. My goal is to be able to issue four commands: Turn right, turn left, move forward, and stop.

I am aware of hand motor imagery detection using Common Spatial Patterns to distinguish between right hand, left hand, or neutral activity. I have been able to run this scenario successfully.

I am also aware of foot motor imagery detection using a surface Laplacian filter to detect beta rebound (although I have yet to find solid documentation supporting this method). If I understand this method correctly, it only supports one command. Furthermore, I have not yet attempted it.

If I am able to combine the above scenarios, as I believe I can through proper electrode placement and data acquisition like channel selection and filtering (please correct me if I'm wrong), I would only be able to issue 3 commands. I am still in need of a fourth command.

I have read a good deal of papers combining MI EEG with EMG to establish the amount of necessary commands, so my inclination is to focus on including EMG feature detection. Could I use jaw clenching as my fourth feature? Also, I would like to avoid using SSVEP or P300 as the goal of this project is to allow the user to keep their eyes on their surroundings when operating the chair.

I am not looking for any incredibly detailed answers! If anyone with experience with a similar project could tell me if they think I'm moving in the right direction or not, I'd really appreciate it.

Thank you!

jtlindgren
Posts: 778
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: BCI Controller for Semi-Autonomous Wheelchair

Post by jtlindgren »

Hi,

its definitely possible to do at least a subset of what you want with openvibe. If you want to use EMG and the two signals well synchronized, the best is if you can plug in the EMG sensor to the same amplifier you get the EEG from. I don't have the time to write a treatise how to do all what you want, but to answer a specific question in the very beginning, in OV of course in the online scenario you'll have csp and classifier operating in the same one. In training, two scenarios are normally used. It might be possible to put them in one so that the classifier training was delayed and data buffered until the CSP solution has been computed, but I haven't tried it.

EEG-based wheelchair control has been often attempted. For example my old colleague Marsel Mano has made some work on it,

http://www.ijicic.org/ijicic-13-02013.pdf

In my opinion the biggest challenge is not related to openvibe per ce, but to the quality of the control you can currently obtain from EEG by any known means. If one can't obtain good classification accuracies (bitrate) using motor imagery without wheelchairs involved, putting a wheelchair in doesn't solve that. It just adds complexity.

Edit: removed some unnecessary politics

My 3 cents,
Jussi

traviskv
Posts: 15
Joined: Thu Aug 02, 2018 12:42 am
Location: USA

Re: BCI Controller for Semi-Autonomous Wheelchair

Post by traviskv »

Thank you for your response and input, Jussi.

I'm aware of the lower data transfer issues with noninvasive EEG, but I'd like to see how far I can push it.

Does the EMG/EEG combination sound more feasible than a combination of right/left hand and foot motor imagery? The idea I have in mind is to use EMG or foot motor imagery detection to activate a toggle switch.

I am using a BCI controller with 14 working electrodes, so I would be devoting a larger portion of those to EEG detection and a smaller portion to EMG.

Thank you!
-Travis

Post Reply