Language model applications to spelling with Brain-Computer Interfaces.

Within the Ambient Assisted Living (AAL) community, Brain-Computer Interfaces (BCIs) have raised great hopes as they provide alternative communication means for persons with disabilities bypassing the need for speech and other motor activities. Although significant advancements have been realized in the last decade, applications of language models (e.g., word prediction, completion) have only recently started to appear in BCI systems. The main goal of this article is to review the language model applications that supplement non-invasive BCI-based communication systems by discussing their potential and limitations, and to discern future trends. First, a brief overview of the most prominent BCI spelling systems is given, followed by an in-depth discussion of the language models applied to them. These language models are classified according to their functionality in the context of BCI-based spelling: the static/dynamic nature of the user interface, the use of error correction and predictive spelling, and the potential to improve their classification performance by using language models. To conclude, the review offers an overview of the advantages and challenges when implementing language models in BCI-based communication systems when implemented in conjunction with other AAL technologies.

Single-Trial ERP Component Analysis Using a Spatiotemporal LCMV Beamformer.

GOAL: For statistical analysis of event-related potentials (ERPs), there are convincing arguments against averaging across stimuli or subjects. Multivariate filters can be used to isolate an ERP component of interest without the averaging procedure. However, we would like to have certainty that the output of the filter accurately represents the component. METHODS: We extended the linearly constrained minimum variance (LCMV) beamformer, which is traditionally used as a spatial filter for source localization, to be a flexible spatiotemporal filter for estimating the amplitude of ERP components in sensor space. In a comparison study on both simulated and real data, we demonstrated the strengths and weaknesses of the beamformer as well as a range of supervised learning approaches. RESULTS: In the context of measuring the amplitude of a specific ERP component on a single-trial basis, we found that the spatiotemporal LCMV beamformer is a filter that accurately captures the component of interest, even in the presence of both structured noise (e.g., other overlapping ERP components) and unstructured noise (e.g., ongoing brain activity and sensor noise). CONCLUSION: The spatiotemporal LCMV beamformer method provides an accurate and intuitive way to conduct analysis of a known ERP component, without averaging across trials or subjects. SIGNIFICANCE: Eliminating averaging allows us to test more detailed hypotheses and apply more powerful statistical models. For example, it allows the usage of multilevel regression models that can incorporate between subject/stimulus variation as random effects, test multiple effects simultaneously, and control confounding effects by partial regression.

Sampled sinusoidal stimulation profile and multichannel fuzzy logic classification for monitor-based phase-coded SSVEP brain-computer interfacing.

OBJECTIVE: The performance and usability of brain-computer interfaces (BCIs) can be improved by new paradigms, stimulation methods, decoding strategies, sensor technology etc. In this study we introduce new stimulation and decoding methods for electroencephalogram (EEG)-based BCIs that have targets flickering at the same frequency but with different phases. APPROACH: The phase information is estimated from the EEG data, and used for target command decoding. All visual stimulation is done on a conventional (60-Hz) LCD screen. Instead of the 'on/off' visual stimulation, commonly used in phase-coded BCI, we propose one based on a sampled sinusoidal intensity profile. In order to fully exploit the circular nature of the evoked phase response, we introduce a filter feature selection procedure based on circular statistics and propose a fuzzy logic classifier designed to cope with circular information from multiple channels jointly. MAIN RESULTS: We show that the proposed visual stimulation enables us not only to encode more commands under the same conditions, but also to obtain EEG responses with a more stable phase. We also demonstrate that the proposed decoding approach outperforms existing ones, especially for the short time windows used. SIGNIFICANCE: The work presented here shows how to overcome some of the limitations of screen-based visual stimulation. The superiority of the proposed decoding approach demonstrates the importance of preserving the circularity of the data during the decoding stage.

Multichannel decoding for phase-coded SSVEP brain-computer interface.

We propose a complex-valued multilayer feedforward neural network classifier for decoding of phase-coded information from steady-state visual evoked potentials. To optimize the performance of the classifier we supply it with two filter-based feature selection strategies. The proposed approaches could be used for a phase-coded brain-computer interface, enabling to encode several targets using only one stimulation frequency. The proposed classifier is a multichannel one, which distinguishes our approach from the existing single-channel ones. We show that the proposed approach outperforms others in terms of accuracy and length of the data segments used for decoding. We show that the decoding based on one optimally selected channel yields an inferior performance compared to the one based on several features, which supports our argument for a multichannel approach.

Comparison of classification methods for P300 brain-computer interface on disabled subjects.

We report on tests with a mind typing paradigm based on a P300 brain-computer interface (BCI) on a group of amyotrophic lateral sclerosis (ALS), middle cerebral artery (MCA) stroke, and subarachnoid hemorrhage (SAH) patients, suffering from motor and speech disabilities. We investigate the achieved typing accuracy given the individual patient's disorder, and how it correlates with the type of classifier used. We considered 7 types of classifiers, linear as well as nonlinear ones, and found that, overall, one type of linear classifier yielded a higher classification accuracy. In addition to the selection of the classifier, we also suggest and discuss a number of recommendations to be considered when building a P300-based typing system for disabled subjects.

Learning eye vergence control from a distributed disparity representation.

We present two neural models for vergence angle control of a robotic head, a simplified and a more complex one. Both models work in a closed-loop manner and do not rely on explicitly computed disparity, but extract the desired vergence angle from the post-processed response of a population of disparity tuned complex cells, the actual gaze direction and the actual vergence angle. The first model assumes that the gaze direction of the robotic head is orthogonal to its baseline and the stimulus is a frontoparallel plane orthogonal to the gaze direction. The second model goes beyond these assumptions, and operates reliably in the general case where all restrictions on the orientation of the gaze, as well as the stimulus position, type and orientation, are dropped.