Riemannian geometry for EEG-based brain-computer interfaces; a primer and a review

Congedo, Marco; Barachant, Alexandre; Bhatia, Rajendra

doi:10.1080/2326263x.2017.1297192

Cited by 325 publications

(314 citation statements)

References 103 publications

Supporting

Mentioning

285

Contrasting

Order By: Relevance

“…models robust to temporal variability of neural correlates, increased noise, interaction of neural processes. Some promising approaches include the use of novel features based on the connectivity across brain areas [6], [7], [15], [53]- [55] or the covariance across channels [56], deep learning [10], [57], as well as techniques for robust decoder training using limited samples such as transfer learning or semi-supervised approaches [9], [40], [58]- [61]. A recent review on current trends for EEG decoding in BMI applications can be found in reference [62].…”

Section: Discussionmentioning

confidence: 99%

Decoding Neural Correlates of Cognitive States to Enhance Driving Experience

Chavarriaga

Uscumlic

Zhang

et al. 2018

IEEE Trans. Emerg. Top. Comput. Intell.

View full text Add to dashboard Cite

Modern cars can support their drivers by assessing and performing autonomously different driving maneuvers, based on information gathered by in-car sensors. We propose that brain machine interfaces (BMIs) can provide complementary information that can ease the interaction with intelligent cars in order to enhance the driving experience. In our approach, the human remains in control, while a BMI is used to monitor the driver's cognitive state and use that information to modulate the assistance provided by the intelligent car. In this review, we gather our proof-of-concept studies demonstrating the feasibility of decoding electroencephalography (EEG) correlates of upcoming actions and those reflecting whether the decisions of driving assistant systems are in-line with the driver intentions. Experimental results while driving both simulated and real cars consistently showed neural signatures of anticipation, movement preparation and error processing. Remarkably, despite the increased noise inherent to real scenarios, these signals can be decoded on a single-trial basis, reflecting some of the cognitive processes that take place while driving. However, moderate decoding performance compared to the controlled experimental BMI paradigms indicate there exists room for improvement of the machine learning methods typically used in the state-ofthe-art BMIs. We foresee that fusion of neural correlates with information extracted from other physiological measures; e.g. eye movements or electromyography (EMG) as well as contextual information gathered by in-car sensors will allow intelligent cars to provide timely and tailored assistance only if it is required; thus keeping the user in the loop and allowing him to fully enjoy the driving experience.

show abstract

Section: Discussionmentioning

confidence: 99%

Decoding Neural Correlates of Cognitive States to Enhance Driving Experience

Chavarriaga

Uscumlic

Zhang

et al. 2018

IEEE Trans. Emerg. Top. Comput. Intell.

View full text Add to dashboard Cite

show abstract

“…The first 'Euclidian' approach consisted in vectorizing the MEG data (each trial being a 306 MEG channels * 150 time samples vector). The second, 'Riemannian' approach consisted in applying 2 Xdawn spatial filters (Rivet et al, 2009), and projecting the resulting event related field meta-covariance data (Congedo 2017) into the tangent space with pyRiemann. For both approaches, we estimated the distances between neural patterns as well as the confusion matrices of the logistic regression (the % of trials predictions for each class).…”

Section: Methodsmentioning

confidence: 99%

Riemannian Geometry Boosts Representational Similarity Analyses of Dense Neural Time Series

Barachant

2017

Preprint

Self Cite

View full text Add to dashboard Cite

Representational similarity analysis (RSA) is a popular technique to estimate the structure of mental representations from neuroimaging data. However, RSA can be difficult to estimate for neural time series, where mental representations may be distributed in a highly dimensional space. Here, we show that RSA can be efficiently estimated from dense neural time series using Riemannian geometry. Using a public magnetoencephalography dataset, we decoded 24 classes from the brain evoked responses to 720 visual stimuli. RSA estimated from the confusion matrices of a standard regularized logistic regression achieved an average decoding accuracy of 23% (chance=4%). Our approach based on spatial filtering and Riemannian geometry nearly doubled this score with an average 42% decoding accuracy. Finally, our results revealed how RSA becomes ill-conceived when it derives from confusion matrices of highly accurate multivariate pattern classifications. Instead, we propose to directly estimate RSA from Riemannian metrics without fitting a multivariate pattern classifier. Overall, our approach, based on Riemannian geometry provides a principled and efficient basis to study the structure of mental representations from highly dimensional neural time series.

show abstract

“…Moreover, [13], [14] have reported an adaptive P300 BCI that does not require calibration. This makes BCI technology more suitable for the general public, since avoiding the need for calibration is a key feature in providing a plug-and-play technology [15]. For the other main bottleneck, the encumbrance and cost of EEG hardware, the readiness of BCI technology is a matter of time, since both the bulkiness and the cost are currently being rapidly reduced (e.g., OpenBCI, New York, US).…”

Section: Introductionmentioning

confidence: 99%

“…interface we developed runs independently on the smartphone and not on a PC. We also implement a robust BCI based on Riemannian geometry, meeting the functional requirements for BCIs of [15]. Finally, we correct the tagging latency in VR and PC, which has never been done before, although it must be corrected to compare the ERPs in the two conditions.…”

Section: Introductionmentioning

confidence: 99%

A Comparison of Mobile VR Display Running on an Ordinary Smartphone With Standard PC Display for P300-BCI Stimulus Presentation

Cattan

Andreev

Mendoza³

et al. 2021

IEEE Trans. Games

Self Cite

View full text Add to dashboard Cite

A brain-computer interface (BCI) based on electroencephalography (EEG) is a promising technology for enhancing virtual reality (VR) applications-in particular, for gaming. We focus on the so-called P300-BCI, a stable and accurate BCI paradigm relying on the recognition of a positive eventrelated potential (ERP) occurring in the EEG about 300 ms poststimulation. We implemented a basic version of such a BCI displayed on an ordinary and affordable smartphone-based headmounted VR device: that is, a mobile and passive VR system (with no electronic components beyond the smartphone). The mobile phone performed the stimuli presentation, EEG synchronization (tagging) and feedback display. We compared the ERPs and the accuracy of the BCI on the VR device with a traditional BCI running on a personal computer (PC). We also evaluated the impact of subjective factors on the accuracy. The study was within-subjects, with 21 participants and one session in each modality. No significant difference in BCI accuracy was found between the PC and VR systems, although the P200 ERP was significantly wider and larger in the VR system as compared to the PC system.

show abstract

Riemannian geometry for EEG-based brain-computer interfaces; a primer and a review

Cited by 325 publications

References 103 publications

Decoding Neural Correlates of Cognitive States to Enhance Driving Experience

Decoding Neural Correlates of Cognitive States to Enhance Driving Experience

Riemannian Geometry Boosts Representational Similarity Analyses of Dense Neural Time Series

A Comparison of Mobile VR Display Running on an Ordinary Smartphone With Standard PC Display for P300-BCI Stimulus Presentation

Contact Info

Product

Resources

About