Developing a 3- to 6-state EEG-based brain-computer interface for a robotic manipulator control

Abstract

Abstract-Recent developments in BCI techniques have demonstrated high-performance control of robotic prosthetic systems primarily via invasive methods. In this work we develop an electroencephalography (EEG) based noninvasive BCI system that can be used for a similar, albeit lower-speed robotic control, and a signal processing system for detecting user's mental intent from EEG data based on up to 6-state motor-imagery BCI communication paradigm. We examine the performance of that system on experimental data collected from 12 healthy participants and analyzed offline. We show that our EEG BCI system can correctly identify different motor imageries in EEG data with high accuracy: 3 out of 12 participants achieved accuracy of 6-state communication in 80-90% range, while 2 participants could not achieve a satisfactory accuracy. We further implement an online BCI system for control of a virtual 3 degree-of-freedom prosthetic manipulator and test it with our 3 best participants. The participants' ability to control the BCI is quantified by using the percentage of successfully completed BCI tasks, the time required to complete a task, and the error rate. 2 participants were able to successfully complete 100% of the test tasks, demonstrating on average the error rate of 80% and requiring 5-10 seconds to execute a manipulator move. 1 participant failed to demonstrate a satisfactory performance in online trials. Our results lay a foundation for further development of EEG BCI-based robotic assistive systems and demonstrate that EEG-based BCI may be feasible for robotic control by paralyzed and immobilized individuals.Index Terms -Brain Machine interfaces, electroencephalography, neural prosthetics.

A large electroencephalographic motor imagery dataset for electroencephalographic brain computer interfaces

“…In this work, we publish an EEG BCI dataset (Data Citation 1) collected during the development of a slow cortical potentials MI EEG BCI in Mishchenko et al 40 . The dataset contains 60 hours of EEG BCI recordings across 75 recording sessions of 13 participants, 60,000 mental imageries, and 4 BCI interaction paradigms, with multiple recording sessions and paradigms of the same individuals.…”

“…The recording sessions in the dataset with the “Inter” identifier were collected using custom software that we call interactive interface, and is described in greater details in Mishchenko et al 40 . The key difference is that Matlab software directly interfaced with Neurofax software to accesses the EEG data in real time to enable real-time feedback to be given in a BCI system.…”

“…ERP is a classical way of representing EEG responses 41–47 . We also carried out a classification of the mental imageries included in each data file using an SVM machine learning algorithm 48–53 and an approach detailed in Mishchenko et al 40 . The results of that validation are reported in Figs.…”

“…We follow the methodology described in Mishchenko et al 40 for mental imagery classification study. Interested readers are requested to refer to that manuscript for details.…”

Empirical Mode Decomposition and a Bidirectional LSTM Architecture Used to Decode Individual Finger MI-EEG Signals

“…Regularly, digit from 1 to up 5 appears on each finger to guide the subject under tests to imagine the up and down finger flexion movements. At the same time, the EEG-1200 equipment captures and registers the corresponding signals via Neurofax recording software [34].…”