Non-stationarity and Inter-subject variability of EEG characteristics in the Context of BCI Development

In a vision of a perfect brain-computer interface (BCI), a user would be able to use the system instantly without the need for a subject-specific calibration, and the performance would remain stable and not deteriorate over time. However, this remains a vision, due to two characteristics of the electroencephalography (EEG) signals: non-stationarity and inter-subject variability. Inter-subject variability describes the fact that the EEG of each person is different and for sufficient BCI communication, the BCI needs to be calibrated separately for each user. Non-stationarity describes a change over time of the EEG signals leading to a decrease in BCI performance with prolonged use. In an approach to better understand these issues, we analyzed the event-related potentials (ERP) and spectral EEG data of 23 subjects in terms of these characteristics. We found that both issues highly affect the data, but we were able to identify a method that nearly eliminates non-stationarity, whereas inter-subject variability remains a major issue that needs to be further addressed.

[1]  Wolfgang Rosenstiel,et al.  Online Adaptation of a c-VEP Brain-Computer Interface(BCI) Based on Error-Related Potentials and Unsupervised Learning , 2012, PloS one.

[2]  Zhanpeng Jin,et al.  Brainprint: Assessing the uniqueness, collectability, and permanence of a novel method for ERP biometrics , 2015, Neurocomputing.

[3]  Wolfgang Rosenstiel,et al.  Principal component based covariate shift adaption to reduce non-stationarity in a MEG-based brain-computer interface , 2012, EURASIP J. Adv. Signal Process..

[4]  M Congedo,et al.  A review of classification algorithms for EEG-based brain–computer interfaces , 2007, Journal of neural engineering.

[5]  Benjamin Schrauwen,et al.  A P300 BCI for the Masses: Prior Information Enables Instant Unsupervised Spelling , 2012, NIPS.

[6]  Wolfgang Rosenstiel,et al.  Online use of error-related potentials in healthy users and people with severe motor impairment increases performance of a P300-BCI , 2012, Clinical Neurophysiology.

[7]  Klaus-Robert Müller,et al.  Integrating dynamic stopping, transfer learning and language models in an adaptive zero-training ERP speller , 2014, Journal of neural engineering.

[8]  G. Pfurtscheller,et al.  Event-related cortical desynchronization detected by power measurements of scalp EEG. , 1977, Electroencephalography and clinical neurophysiology.

[9]  Wolfgang Rosenstiel,et al.  EEG-based prediction of cognitive workload induced by arithmetic: a step towards online adaptation in numerical learning , 2016 .

[10]  Ramaswamy Palaniappan,et al.  Two-Stage Biometric Authentication Method Using Thought Activity Brain Waves , 2008, Int. J. Neural Syst..

[11]  Wolfgang Rosenstiel,et al.  Adaptive SVM-Based Classification Increases Performance of a MEG-Based Brain-Computer Interface (BCI) , 2012, ICANN.

[12]  Dongrui Wu,et al.  Collaborative Filtering for Brain-Computer Interaction Using Transfer Learning and Active Class Selection , 2013, PloS one.

[13]  J. Burg.  THE RELATIONSHIP BETWEEN MAXIMUM ENTROPY SPECTRA AND MAXIMUM LIKELIHOOD SPECTRA , 1972 .