BCI meeting 2005-workshop on BCI signal processing: feature extraction and translation

SpeechBrain-MOABB: An open-source Python library for benchmarking deep neural networks applied to EEG signals

Deep learning has revolutionized EEG decoding, showcasing its ability to outperform traditional machine learning models. However, unlike other fields, EEG decoding lacks comprehensive open-source libraries dedicated to neural networks. Existing tools (MOABB and braindecode) prevent the creation of robust and complete decoding pipelines, as they lack support for hyperparameter search across the entire pipeline, and are sensitive to fluctuations in results due to network random initialization. Furthermore, the absence of a standardized experimental protocol exacerbates the reproducibility crisis in the field. To address these limitations, we introduce SpeechBrain-MOABB, a novel open-source toolkit carefully designed to facilitate the development of a comprehensive EEG decoding pipeline based on deep learning. SpeechBrain-MOABB incorporates a complete experimental protocol that standardizes critical phases, such as hyperparameter search and model evaluation. It natively supports multi-step hyperparameter search for finding the optimal hyperparameters in a high-dimensional space defined by the entire pipeline, and multi-seed training and evaluation for obtaining performance estimates robust to the variability caused by random initialization. SpeechBrain-MOABB outperforms other libraries, including MOABB and braindecode, with accuracy improvements of 14.9% and 25.2% (on average), respectively. By enabling easy-to-use and easy-to-share decoding pipelines, our toolkit can be exploited by neuroscientists for decoding EEG with neural networks in a replicable and trustworthy way.

A causal perspective on brainwave modeling for brain-computer interfaces

Objective. Machine learning (ML) models have opened up enormous opportunities in the field of brain-computer Interfaces (BCIs). Despite their great success, they usually face severe limitations when they are employed in real-life applications outside a controlled laboratory setting.Approach. Mixing causal reasoning, identifying causal relationships between variables of interest, with brainwave modeling can change one's viewpoint on some of these major challenges which can be found in various stages in the ML pipeline, ranging from data collection and data pre-processing to training methods and techniques.Main results. In this work, we employ causal reasoning and present a framework aiming to breakdown and analyze important challenges of brainwave modeling for BCIs.Significance. Furthermore, we present how general ML practices as well as brainwave-specific techniques can be utilized and solve some of these identified challenges. And finally, we discuss appropriate evaluation schemes in order to measure these techniques' performance and efficiently compare them with other methods that will be developed in the future.

BrainWave-Scattering Net: a lightweight network for EEG-based motor imagery recognition

Objective.Brain-computer interfaces (BCIs) enable a direct communication of the brain with the external world, using one's neural activity, measured by electroencephalography (EEG) signals. In recent years, convolutional neural networks (CNNs) have been widely used to perform automatic feature extraction and classification in various EEG-based tasks. However, their undeniable benefits are counterbalanced by the lack of interpretability properties as well as the inability to perform sufficiently when only limited amount of training data is available.Approach.In this work, we introduce a novel, lightweight, fully-learnable neural network architecture that relies on Gabor filters to delocalize EEG signal information into scattering decomposition paths along frequency and slow-varying temporal modulations.Main results.We utilize our network in two distinct modeling settings, for building either a generic (training across subjects) or a personalized (training within a subject) classifier.Significance.In both cases, using two different publicly available datasets and one in-house collected dataset, we demonstrate high performance for our model with considerably less number of trainable parameters as well as shorter training time compared to other state-of-the-art deep architectures. Moreover, our network demonstrates enhanced interpretability properties emerging at the level of the temporal filtering operation and enables us to train efficient personalized BCI models with limited amount of training data.

Convolutional Neural Network with a Topographic Representation Module for EEG-Based Brain-Computer Interfaces

Convolutional neural networks (CNNs) have shown great potential in the field of brain-computer interfaces (BCIs) due to their ability to directly process raw electroencephalogram (EEG) signals without artificial feature extraction. Some CNNs have achieved better classification accuracy than that of traditional methods. Raw EEG signals are usually represented as a two-dimensional (2-D) matrix composed of channels and time points, ignoring the spatial topological information of electrodes. Our goal is to make a CNN that takes raw EEG signals as inputs have the ability to learn spatial topological features and improve its classification performance while basically maintaining its original structure. We propose an EEG topographic representation module (TRM). This module consists of (1) a mapping block from raw EEG signals to a 3-D topographic map and (2) a convolution block from the topographic map to an output with the same size as the input. According to the size of the convolutional kernel used in the convolution block, we design two types of TRMs, namely TRM-(5,5) and TRM-(3,3). We embed the two TRM types into three widely used CNNs (ShallowConvNet, DeepConvNet and EEGNet) and test them on two publicly available datasets (the Emergency Braking During Simulated Driving Dataset (EBDSDD) and the High Gamma Dataset (HGD)). Results show that the classification accuracies of all three CNNs are improved on both datasets after using the TRMs. With TRM-(5,5), the average classification accuracies of DeepConvNet, EEGNet and ShallowConvNet are improved by 6.54%, 1.72% and 2.07% on the EBDSDD and by 6.05%, 3.02% and 5.14% on the HGD, respectively; with TRM-(3,3), they are improved by 7.76%, 1.71% and 2.17% on the EBDSDD and by 7.61%, 5.06% and 6.28% on the HGD, respectively. We improve the classification performance of three CNNs on both datasets through the use of TRMs, indicating that they have the capability to mine spatial topological EEG information. More importantly, since the output of a TRM has the same size as the input, CNNs with raw EEG signals as inputs can use this module without changing their original structures.

Mental Flow Estimation Through Wearable EEG

Flow is a mental state experienced during holistic involvement in a certain task, and it is a factor that promotes motivation, development, and performance. A reliable and objective estimation of the flow is essential for moving away from the traditional self-reporting subjective questionnaires, and for developing closed-loop human-computer interfaces. In this study, we recorded EEG and pupil dilation in a cohort of participants solving arithmetic problems. In particular, the EEG activity was acquired with a prototype of a commercial headset from Logitech with nine dry electrodes incorporated in a pair of over-ear headphones. The difficulty of the tasks was adapted to induce mental Boredom, Flow and Overload, corresponding to too easy, optimal and too challenging tasks, respectively. Results indicated statistically significant differences between all pairs of conditions for the pupil dilation, as well as for the EEG activity for the electrodes in the ear-pads. Furthermore, we built a predictive model that estimated the mental state of the user from their EEG data with 65% accuracy.

MCGNet+: an improved motor imagery classification based on cosine similarity

It has been a challenge for solving the motor imagery classification problem in the brain informatics area. Accuracy and efficiency are the major obstacles for motor imagery analysis in the past decades since the computational capability and algorithmic availability cannot satisfy complex brain signal analysis. In recent years, the rapid development of machine learning (ML) methods has empowered people to tackle the motor imagery classification problem with more efficient methods. Among various ML methods, the Graph neural networks (GNNs) method has shown its efficiency and accuracy in dealing with inter-related complex networks. The use of GNN provides new possibilities for feature extraction from brain structure connection. In this paper, we proposed a new model called MCGNet⁺, which improves the performance of our previous model MutualGraphNet. In this latest model, the mutual information of the input columns forms the initial adjacency matrix for the cosine similarity calculation between columns to generate a new adjacency matrix in each iteration. The dynamic adjacency matrix combined with the spatial temporal graph convolution network (ST-GCN) has better performance than the unchanged matrix model. The experimental results indicate that MCGNet⁺ is robust enough to learn the interpretable features and outperforms the current state-of-the-art methods.

Progress in Brain Computer Interface: Challenges and Opportunities

Brain computer interfaces (BCI) provide a direct communication link between the brain and a computer or other external devices. They offer an extended degree of freedom either by strengthening or by substituting human peripheral working capacity and have potential applications in various fields such as rehabilitation, affective computing, robotics, gaming, and neuroscience. Significant research efforts on a global scale have delivered common platforms for technology standardization and help tackle highly complex and non-linear brain dynamics and related feature extraction and classification challenges. Time-variant psycho-neurophysiological fluctuations and their impact on brain signals impose another challenge for BCI researchers to transform the technology from laboratory experiments to plug-and-play daily life. This review summarizes state-of-the-art progress in the BCI field over the last decades and highlights critical challenges.

A robust and reliable online P300-based BCI system using Emotiv EPOC + headset

Brain-computer interface (BCI) system aims to enable interaction with people and therefore the environment without muscular activation, using changes in brain signals due to the execution of cognitive tasks. The target of the presented work is to investigate the power of Emotiv EPOC + headset to detect and record the P300 wave. Moreover, the effect of preprocessing the acquired signal was studied. Five participants were asked to attend different sessions to an equivalent 6x6 matrix while the rows and columns were randomly flashed at a rate of 200 ms. The acquired EEG data were sent wirelessly to OpenViBE software, which is employed to run the P300 speller. Two classification methods were tried: Linear discriminate analysis (LDA) and support vector machine (SVM). The capability of the headset to detect the P300 signals is proven by the results. Additionally, results show that participants reached accuracy up to 90 and 70% after only two training sessions for Linear discriminate analysis (LDA) and support vector machine (SVM) classifiers, respectively. The significance of this work is to demonstrate that such a portable and affordable headset might be useful to design and implement a robust and reliable online P300-based BCI system.

Subject-independent decoding of affective states using functional near-infrared spectroscopy

Affective decoding is the inference of human emotional states using brain signal measurements. This approach is crucial to develop new therapeutic approaches for psychiatric rehabilitation, such as affective neurofeedback protocols. To reduce the training duration and optimize the clinical outputs, an ideal clinical neurofeedback could be trained using data from an independent group of volunteers before being used by new patients. Here, we investigated if this subject-independent design of affective decoding can be achieved using functional near-infrared spectroscopy (fNIRS) signals from frontal and occipital areas. For this purpose, a linear discriminant analysis classifier was first trained in a dataset (49 participants, 24.65±3.23 years) and then tested in a completely independent one (20 participants, 24.00±3.92 years). Significant balanced accuracies between classes were found for positive vs. negative (64.50 ± 12.03%, p<0.01) and negative vs. neutral (68.25 ± 12.97%, p<0.01) affective states discrimination during a reactive block consisting in viewing affective-loaded images. For an active block, in which volunteers were instructed to recollect personal affective experiences, significant accuracy was found for positive vs. neutral affect classification (71.25 ± 18.02%, p<0.01). In this last case, only three fNIRS channels were enough to discriminate between neutral and positive affective states. Although more research is needed, for example focusing on better combinations of features and classifiers, our results highlight fNIRS as a possible technique for subject-independent affective decoding, reaching significant classification accuracies of emotional states using only a few but biologically relevant features.

Single-Option P300-BCI Performance Is Affected by Visual Stimulation Conditions

The P300 paradigm is one of the most promising techniques for its robustness and reliability in Brain-Computer Interface (BCI) applications, but it is not exempt from shortcomings. The present work studied single-trial classification effectiveness in distinguishing between target and non-target responses considering two conditions of visual stimulation and the variation of the number of symbols presented to the user in a single-option visual frame. In addition, we also investigated the relationship between the classification results of target and non-target events when training and testing the machine-learning model with datasets containing different stimulation conditions and different number of symbols. To this end, we designed a P300 experimental protocol considering, as conditions of stimulation: the color highlighting or the superimposing of a cartoon face and from four to nine options. These experiments were carried out with 19 healthy subjects in 3 sessions. The results showed that the Event-Related Potentials (ERP) responses and the classification accuracy are stronger with cartoon faces as stimulus type and similar irrespective of the amount of options. In addition, the classification performance is reduced when using datasets with different type of stimulus, but it is similar when using datasets with different the number of symbols. These results have a special connotation for the design of systems, in which it is intended to elicit higher levels of evoked potentials and, at the same time, optimize training time.

Interpretable and lightweight convolutional neural network for EEG decoding: Application to movement execution and imagination

Convolutional neural networks (CNNs) are emerging as powerful tools for EEG decoding: these techniques, by automatically learning relevant features for class discrimination, improve EEG decoding performances without relying on handcrafted features. Nevertheless, the learned features are difficult to interpret and most of the existing CNNs introduce many trainable parameters. Here, we propose a lightweight and interpretable shallow CNN (Sinc-ShallowNet), by stacking a temporal sinc-convolutional layer (designed to learn band-pass filters, each having only the two cut-off frequencies as trainable parameters), a spatial depthwise convolutional layer (reducing channel connectivity and learning spatial filters tied to each band-pass filter), and a fully-connected layer finalizing the classification. This convolutional module limits the number of trainable parameters and allows direct interpretation of the learned spectral-spatial​ features via simple kernel visualizations. Furthermore, we designed a post-hoc gradient-based technique to enhance interpretation by identifying the more relevant and more class-specific features. Sinc-ShallowNet was evaluated on benchmark motor-execution and motor-imagery datasets and against different design choices and training strategies. Results show that (i) Sinc-ShallowNet outperformed a traditional machine learning algorithm and other CNNs for EEG decoding; (ii) The learned spectral-spatial features matched well-known EEG motor-related activity; (iii) The proposed architecture performed better with a larger number of temporal kernels still maintaining a good compromise between accuracy and parsimony, and with a trialwise rather than a cropped training strategy. In perspective, the proposed approach, with its interpretative capacity, can be exploited to investigate cognitive/motor aspects whose EEG correlates are yet scarcely known, potentially characterizing their relevant features.

MULTIPLE CLASSIFICATION TECHNIQUES TOWARD A ROBUST AND RELIABLE P300 BCI SYSTEM

A brain–computer interface (BCI) can be used for people with severe physical disabilities such as ALS or amyotrophic lateral sclerosis. BCI can allow these individuals to communicate again by creating a new communication channel directly from the brain to an output device. BCI technology can allow paralyzed people to share their intent with others, and thereby demonstrate that direct communication from the brain to the external world is possible and that it might serve useful functions. BCI systems include machine learning algorithms (MLAs). Their performance depends on the feature extraction and classification techniques employed. In this paper, we propose a system to exploit the P300 signal in the brain, a positive deflection in event-related potentials. The P300 signal can be incorporated into a spelling device. There are two benefits behind this kind of research. First of all, this work presents the research status and the advantages of communication via a BCI system, especially the P300 BCI system for disordered people, and the related literature review is presented. Secondly, the paper discusses the performance of different machine learning algorithms. Two different datasets are presented: the first dataset 2004 and the second dataset 2019. A preprocessing step is introduced to the subjects in both datasets first to extract the important features before applying the proposed machine learning methods: linear discriminant analysis (LDA I and LDA II), support vector machine (SVM I, SVM II, SVM III, and SVM IV), linear regression (LREG), Bayesian linear discriminant analysis (BLDA), and twin support vector machine (TSVM). By comparing the performance of the different machine learning systems, in the first dataset it is found that BLDA and SVMIV classifiers yield the highest performance for both subjects “A” and “B”. BLDA yields 98% and 66% for 15th and 5th sequences, respectively, whereas SVMIV yields 98% and 54.4% for 15th and 5th sequences, respectively. While in the second dataset, it is obvious that BLDA classifier yields the highest performance for both subjects “1” and “2”, it achieves 90.115%. The paper summarizes the P300 BCI system for the two introduced datasets. It discusses the proposed system, compares the classification methods performances, and considers some aspects for the future work to be handled. The results show high accuracy and less computational time which makes the system more applicable for online applications.

A Parallel Multiscale Filter Bank Convolutional Neural Networks for Motor Imagery EEG Classification

OBJECTIVE

Electroencephalogram (EEG) based brain-computer interfaces (BCI) in motor imagery (MI) have developed rapidly in recent years. A reliable feature extraction method is essential because of a low signal-to-noise ratio (SNR) and time-dependent covariates of EEG signals. Because of efficient application in various fields, deep learning has been adopted in EEG signal processing and has obtained competitive results compared with the traditional methods. However, designing and training an end-to-end network to fully extract potential features from EEG signals remains a challenge in MI.

APPROACH

In this study, we propose a parallel multiscale filter bank convolutional neural network (MSFBCNN) for MI classification. We introduce a layered end-to-end network structure, in which a feature-extraction network is used to extract temporal and spatial features. To enhance the transfer learning ability, we propose a network initialization and fine-tuning strategy to train an individual model for inter-subject classification on small datasets. We compare our MSFBCNN with the state-of-the-art approaches on open datasets.

RESULTS

The proposed method has a higher accuracy than the baselines in intra-subject classification. In addition, the transfer learning experiments indicate that our network can build an individual model and obtain acceptable results in inter-subject classification. The results suggest that the proposed network has superior performance, robustness, and transfer learning ability.

A Classification Model for Sensing Human Trust in Machines Using EEG and GSR

Today, intelligent machines interact and collaborate with humans in a way that demands a greater level of trust between human and machine. A first step toward building intelligent machines that are capable of building and maintaining trust with humans is the design of a sensor that will enable machines to estimate human trust level in real time. In this article, two approaches for developing classifier-based empirical trust-sensor models are presented that specifically use electroencephalography and galvanic skin response measurements. Human subject data collected from 45 participants is used for feature extraction, feature selection, classifier training, and model validation. The first approach considers a general set of psychophysiological features across all participants as the input variables and trains a classifier-based model for each participant, resulting in a trust-sensor model based on the general feature set (i.e., a “general trust-sensor model”). The second approach considers a customized feature set for each individual and trains a classifier-based model using that feature set, resulting in improved mean accuracy but at the expense of an increase in training time. This work represents the first use of real-time psychophysiological measurements for the development of a human trust sensor. Implications of the work, in the context of trust management algorithm design for intelligent machines, are also discussed.

EEG-Based Control for Upper and Lower Limb Exoskeletons and Prostheses: A Systematic Review

Electroencephalography (EEG) signals have great impact on the development of assistive rehabilitation devices. These signals are used as a popular tool to investigate the functions and the behavior of the human motion in recent research. The study of EEG-based control of assistive devices is still in early stages. Although the EEG-based control of assistive devices has attracted a considerable level of attention over the last few years, few studies have been carried out to systematically review these studies, as a means of offering researchers and experts a comprehensive summary of the present, state-of-the-art EEG-based control techniques used for assistive technology. Therefore, this research has three main goals. The first aim is to systematically gather, summarize, evaluate and synthesize information regarding the accuracy and the value of previous research published in the literature between 2011 and 2018. The second goal is to extensively report on the holistic, experimental outcomes of this domain in relation to current research. It is systematically performed to provide a wealthy image and grounded evidence of the current state of research covering EEG-based control for assistive rehabilitation devices to all the experts and scientists. The third goal is to recognize the gap of knowledge that demands further investigation and to recommend directions for future research in this area.

Functional near-infrared spectroscopy-based affective neurofeedback: feedback effect, illiteracy phenomena, and whole-connectivity profiles

Background: Affective neurofeedback constitutes a suitable approach to control abnormal neural activities associated with psychiatric disorders and might consequently relief symptom severity. However, different aspects of neurofeedback remain unclear, such as its neural basis, the performance variation, the feedback effect, among others. Aim: First, we aimed to propose a functional near-infrared spectroscopy (fNIRS)-based affective neurofeedback based on the self-regulation of frontal and occipital networks. Second, we evaluated three different feedback approaches on performance: real, fixed, and random feedback. Third, we investigated different demographic, psychological, and physiological predictors of performance. Approach: Thirty-three healthy participants performed a task whereby an amorphous figure changed its shape according to the elicited affect (positive or neutral). During the task, the participants randomly received three different feedback approaches: real feedback, with no change of the classifier output; fixed feedback, keeping the feedback figure unmodified; and random feedback, where the classifier output was multiplied by an arbitrary value, causing a feedback different than expected by the subject. Then, we applied a multivariate comparison of the whole-connectivity profiles according to the affective states and feedback approaches, as well as during a pretask resting-state block, to predict performance. Results: Participants were able to control this feedback system with 70.00 % ± 24.43 % ( p < 0.01 ) of performance during the real feedback trials. No significant differences were found when comparing the average performances of the feedback approaches. However, the whole functional connectivity profiles presented significant Mahalanobis distances ( p ≪ 0.001 ) when comparing both affective states and all feedback approaches. Finally, task performance was positively correlated to the pretask resting-state whole functional connectivity ( r = 0.512 , p = 0.009 ). Conclusions: Our results suggest that fNIRS might be a feasible tool to develop a neurofeedback system based on the self-regulation of affective networks. This finding enables future investigations using an fNIRS-based affective neurofeedback in psychiatric populations. Furthermore, functional connectivity profiles proved to be a good predictor of performance and suggested an increased effort to maintain task control in the presence of feedback distractors.

A fresh look at functional link neural network for motor imagery-based brain-computer interface

BACKGROUND

Artificial neural networks (ANNs) are one of the widely used classifiers in the brain-computer interface (BCI) systems-based on noninvasive electroencephalography (EEG) signals. Among the different ANN architectures, the most commonly applied for BCI classifiers is the multilayer perceptron (MLP). When appropriately designed with optimal number of neuron layers and number of neurons per layer, the ANN can act as a universal approximator. However, due to the low signal-to-noise ratio of EEG signal data, overtraining problem may become an inherent issue, causing these universal approximators to fail in real-time applications.

NEW METHOD

In this study we introduce a higher order neural network, namely the functional link neural network (FLNN) as a classifier for motor imagery (MI)-based BCI systems, to remedy the drawbacks in MLP.

RESULTS

We compare the proposed method with competing classifiers such as linear decomposition analysis, naïve Bayes, k-nearest neighbours, support vector machine and three MLP architectures. Two multi-class benchmark datasets from the BCI competitions are used. Common spatial pattern algorithm is utilized for feature extraction to build classification models.

COMPARISON WITH EXISTING METHOD(S)

FLNN reports the highest average Kappa value over multiple subjects for both the BCI competition datasets, under similarly preprocessed data and extracted features. Further, statistical comparison results over multiple subjects show that the proposed FLNN classification method yields the best performance among the competing classifiers.

CONCLUSIONS

Findings from this study imply that the proposed method, which has less computational complexity compared to the MLP, can be implemented effectively in practical MI-based BCI systems.

Exploration of neural correlates of movement intention based on characterisation of temporal dependencies in electroencephalography

Brain computer interfaces (BCIs) provide a direct communication channel by using brain signals, enabling patients with motor impairments to interact with external devices. Motion intention detection is useful for intuitive movement-based BCI as movement is the fundamental mode of interaction with the environment. The aim of this paper is to investigate the temporal dynamics of brain processes using electroencephalography (EEG) to explore novel neural correlates of motion intention. We investigate the changes in temporal dependencies of the EEG by characterising the decay of autocorrelation during asynchronous voluntary finger tapping movement. The evolution of the autocorrelation function is characterised by its relaxation time, which is used as a robust marker for motion intention. We observed that there was reorganisation of temporal dependencies in EEG during motion intention. The autocorrelation decayed slower during movement intention and faster during the resting state. There was an increase in temporal dependence during movement intention. The relaxation time of the autocorrelation function showed significant (p < 0.05) discrimination between movement and resting state with the mean sensitivity of 78.37 ± 8.83%. The relaxation time provides movement related information that is complementary to the well-known event-related desynchronisation (ERD) by characterising the broad band EEG dynamics which is frequency independent in contrast to ERD. It can also detect motion intention on average 0.51s before the actual movement onset. We have thoroughly compared autocorrelation relaxation time features with ERD in four frequency bands. The relaxation time may therefore, complement the well-known features used in motion-based BCI leading to more robust and intuitive BCI solutions. The results obtained suggest that changes in autocorrelation decay may involve reorganisation of temporal dependencies of brain activity over longer duration during motion intention. This opens the possibilities of investigating further the temporal dynamics of fundamental neural processes underpinning motion intention.

A Biologically Inspired Approach to Frequency Domain Feature Extraction for EEG Classification

Classification of electroencephalogram (EEG) signal is important in mental decoding for brain-computer interfaces (BCI). We introduced a feature extraction approach based on frequency domain analysis to improve the classification performance on different mental tasks using single-channel EEG. This biologically inspired method extracts the most discriminative spectral features from power spectral densities (PSDs) of the EEG signals. We applied our method on a dataset of six subjects who performed five different imagination tasks: (i) resting state, (ii) mental arithmetic, (iii) imagination of left hand movement, (iv) imagination of right hand movement, and (v) imagination of letter “A.” Pairwise and multiclass classifications were performed in single EEG channel using Linear Discriminant Analysis and Support Vector Machines. Our method produced results (mean classification accuracy of 83.06% for binary classification and 91.85% for multiclassification) that are on par with the state-of-the-art methods, using single-channel EEG with low computational cost. Among all task pairs, mental arithmetic versus letter imagination yielded the best result (mean classification accuracy of 90.29%), indicating that this task pair could be the most suitable pair for a binary class BCI. This study contributes to the development of single-channel BCI, as well as finding the best task pair for user defined applications.

Adaptive Spatial Filter Based on Similarity Indices to Preserve the Neural Information on EEG Signals during On-Line Processing

This work presents a new on-line adaptive filter, which is based on a similarity analysis between standard electrode locations, in order to reduce artifacts and common interferences throughout electroencephalography (EEG) signals, but preserving the useful information. Standard deviation and Concordance Correlation Coefficient (CCC) between target electrodes and its correspondent neighbor electrodes are analyzed on sliding windows to select those neighbors that are highly correlated. Afterwards, a model based on CCC is applied to provide higher values of weight to those correlated electrodes with lower similarity to the target electrode. The approach was applied to brain computer-interfaces (BCIs) based on Canonical Correlation Analysis (CCA) to recognize 40 targets of steady-state visual evoked potential (SSVEP), providing an accuracy (ACC) of 86.44 ± 2.81%. In addition, also using this approach, features of low frequency were selected in the pre-processing stage of another BCI to recognize gait planning. In this case, the recognition was significantly (p<0.01) improved for most of the subjects (ACC≥74.79%), when compared with other BCIs based on Common Spatial Pattern, Filter Bank-Common Spatial Pattern, and Riemannian Geometry.

Classifying three imaginary states of the same upper extremity using time-domain features

Brain-computer interface (BCI) allows collaboration between humans and machines. It translates the electrical activity of the brain to understandable commands to operate a machine or a device. In this study, we propose a method to improve the accuracy of a 3-class BCI using electroencephalographic (EEG) signals. This BCI discriminates rest against imaginary grasps and elbow movements of the same limb. This classification task is challenging because imaginary movements within the same limb have close spatial representations on the motor cortex area. The proposed method extracts time-domain features and classifies them using a support vector machine (SVM) with a radial basis kernel function (RBF). An average accuracy of 74.2% was obtained when using the proposed method on a dataset collected, prior to this study, from 12 healthy individuals. This accuracy was higher than that obtained when other widely used methods, such as common spatial patterns (CSP), filter bank CSP (FBCSP), and band power methods, were used on the same dataset. These results are encouraging and the proposed method could potentially be used in future applications including BCI-driven robotic devices, such as a portable exoskeleton for the arm, to assist individuals with impaired upper extremity functions in performing daily tasks.

Effect of Subliminal Lexical Priming on the Subjective Perception of Images: A Machine Learning Approach

The purpose of the study is to examine the effect of subliminal priming in terms of the perception of images influenced by words with positive, negative, and neutral emotional content, through electroencephalograms (EEGs). Participants were instructed to rate how much they like the stimuli images, on a 7-point Likert scale, after being subliminally exposed to masked lexical prime words that exhibit positive, negative, and neutral connotations with respect to the images. Simultaneously, the EEGs were recorded. Statistical tests such as repeated measures ANOVAs and two-tailed paired-samples t-tests were performed to measure significant differences in the likability ratings among the three prime affect types; the results showed a strong shift in the likeness judgment for the images in the positively primed condition compared to the other two. The acquired EEGs were examined to assess the difference in brain activity associated with the three different conditions. The consistent results obtained confirmed the overall priming effect on participants' explicit ratings. In addition, machine learning algorithms such as support vector machines (SVMs), and AdaBoost classifiers were applied to infer the prime affect type from the ERPs. The highest classification rates of 95.0% and 70.0% obtained respectively for average-trial binary classifier and average-trial multi-class further emphasize that the ERPs encode information about the different kinds of primes.

Moving Average Convergence Divergence filter preprocessing for real-time event-related peak activity onset detection : application to fNIRS signals

Real-time solutions for noise reduction and signal processing represent a central challenge for the development of Brain Computer Interfaces (BCI). In this paper, we introduce the Moving Average Convergence Divergence (MACD) filter, a tunable digital passband filter for online noise reduction and onset detection without preliminary learning phase, used in economic markets analysis. MACD performance was tested and benchmarked with other filters using data collected with functional Near Infrared Spectoscopy (fNIRS) during a digit sequence memorization task. This filter has a good performance on filtering and real-time peak activity onset detection, compared to other techniques. Therefore, MACD could be implemented for efficient BCI design using fNIRS.

Real-time state estimation in a flight simulator using fNIRS

Working memory is a key executive function for flying an aircraft. This function is particularly critical when pilots have to recall series of air traffic control instructions. However, working memory limitations may jeopardize flight safety. Since the functional near-infrared spectroscopy (fNIRS) method seems promising for assessing working memory load, our objective is to implement an on-line fNIRS-based inference system that integrates two complementary estimators. The first estimator is a real-time state estimation MACD-based algorithm dedicated to identifying the pilot's instantaneous mental state (not-on-task vs. on-task). It does not require a calibration process to perform its estimation. The second estimator is an on-line SVM-based classifier that is able to discriminate task difficulty (low working memory load vs. high working memory load). These two estimators were tested with 19 pilots who were placed in a realistic flight simulator and were asked to recall air traffic control instructions. We found that the estimated pilot's mental state matched significantly better than chance with the pilot's real state (62% global accuracy, 58% specificity, and 72% sensitivity). The second estimator, dedicated to assessing single trial working memory loads, led to 80% classification accuracy, 72% specificity, and 89% sensitivity. These two estimators establish reusable blocks for further fNIRS-based passive brain computer interface development.

Enhancing brain-machine interface (BMI) control of a hand exoskeleton using electrooculography (EOG)

BACKGROUND

Brain-machine interfaces (BMIs) allow direct translation of electric, magnetic or metabolic brain signals into control commands of external devices such as robots, prostheses or exoskeletons. However, non-stationarity of brain signals and susceptibility to biological or environmental artifacts impede reliable control and safety of BMIs, particularly in daily life environments. Here we introduce and tested a novel hybrid brain-neural computer interaction (BNCI) system fusing electroencephalography (EEG) and electrooculography (EOG) to enhance reliability and safety of continuous hand exoskeleton-driven grasping motions.

FINDINGS

12 healthy volunteers (8 male, mean age 28.1 ± 3.63y) used EEG (condition #1) and hybrid EEG/EOG (condition #2) signals to control a hand exoskeleton. Motor imagery-related brain activity was translated into exoskeleton-driven hand closing motions. Unintended motions could be interrupted by eye movement-related EOG signals. In order to evaluate BNCI control and safety, participants were instructed to follow a visual cue indicating either to move or not to move the hand exoskeleton in a random order. Movements exceeding 25% of a full grasping motion when the device was not supposed to be moved were defined as safety violation. While participants reached comparable control under both conditions, safety was frequently violated under condition #1 (EEG), but not under condition #2 (EEG/EOG).

CONCLUSION

EEG/EOG biosignal fusion can substantially enhance safety of assistive BNCI systems improving their applicability in daily life environments.

Enhanced low-latency detection of motor intention from EEG for closed-loop brain-computer interface applications

In recent years, the detection of voluntary motor intentions from electroencephalogram (EEG) has been used for triggering external devices in closed-loop brain-computer interface (BCI) research. Movement-related cortical potentials (MRCP), a type of slow cortical potentials, have been recently used for detection. In order to enhance the efficacy of closed-loop BCI systems based on MRCPs, a manifold method called Locality Preserving Projection, followed by a linear discriminant analysis (LDA) classifier (LPP-LDA) is proposed in this paper to detect MRCPs from scalp EEG in real time. In an online experiment on nine healthy subjects, LPP-LDA statistically outperformed the classic matched filter approach with greater true positive rate (79 ± 11% versus 68 ± 10%; p = 0.007) and less false positives (1.4 ± 0.8/min versus 2.3 ± 1.1/min; p = 0.016 ). Moreover, the proposed system performed detections with significantly shorter latency (315 ± 165 ms versus 460 ± 123 ms; p = 0.013), which is a fundamental characteristics to induce neuroplastic changes in closed-loop BCIs, following the Hebbian principle. In conclusion, the proposed system works as a generic brain switch, with high accuracy, low latency, and easy online implementation. It can thus be used as a fundamental element of BCI systems for neuromodulation and motor function rehabilitation.

EEG channel selection using particle swarm optimization for the classification of auditory event-related potentials

Brain-machine interfaces (BMI) rely on the accurate classification of event-related potentials (ERPs) and their performance greatly depends on the appropriate selection of classifier parameters and features from dense-array electroencephalography (EEG) signals. Moreover, in order to achieve a portable and more compact BMI for practical applications, it is also desirable to use a system capable of accurate classification using information from as few EEG channels as possible. In the present work, we propose a method for classifying P300 ERPs using a combination of Fisher Discriminant Analysis (FDA) and a multiobjective hybrid real-binary Particle Swarm Optimization (MHPSO) algorithm. Specifically, the algorithm searches for the set of EEG channels and classifier parameters that simultaneously maximize the classification accuracy and minimize the number of used channels. The performance of the method is assessed through offline analyses on datasets of auditory ERPs from sound discrimination experiments. The proposed method achieved a higher classification accuracy than that achieved by traditional methods while also using fewer channels. It was also found that the number of channels used for classification can be significantly reduced without greatly compromising the classification accuracy.

Robust classification of motor imagery EEG signals using statistical time-domain features

The tradeoff between computational complexity and speed, in addition to growing demands for real-time BMI (brain-machine interface) systems, expose the necessity of applying methods with least possible complexity. Willison amplitude (WAMP) and slope sign change (SSC) are two promising time-domain features only if the right threshold value is defined for them. To overcome the drawback of going through trial and error for the determination of a suitable threshold value, modified WAMP and modified SSC are proposed in this paper. Besides, a comprehensive assessment of statistical time-domain features in which their effectiveness is evaluated with a support vector machine (SVM) is presented. To ensure the accuracy of the results obtained by the SVM, the performance of each feature is reassessed with supervised fuzzy C-means. The general assessment shows that every subject had at least one of his performances near or greater than 80%. The obtained results prove that for BMI applications, in which a few errors can be tolerated, these combinations of feature-classifier are suitable. Moreover, features that could perform satisfactorily were selected for feature combination. Combinations of the selected features are evaluated with the SVM, and they could significantly improve the results, in some cases, up to full accuracy.

Characterizing multivariate decoding models based on correlated EEG spectral features

OBJECTIVE

Multivariate decoding methods are popular techniques for analysis of neurophysiological data. The present study explored potential interpretative problems with these techniques when predictors are correlated.

METHODS

Data from sensorimotor rhythm-based cursor control experiments was analyzed offline with linear univariate and multivariate models. Features were derived from autoregressive (AR) spectral analysis of varying model order which produced predictors that varied in their degree of correlation (i.e., multicollinearity).

RESULTS

The use of multivariate regression models resulted in much better prediction of target position as compared to univariate regression models. However, with lower order AR features interpretation of the spectral patterns of the weights was difficult. This is likely to be due to the high degree of multicollinearity present with lower order AR features.

CONCLUSIONS

Care should be exercised when interpreting the pattern of weights of multivariate models with correlated predictors. Comparison with univariate statistics is advisable.

SIGNIFICANCE

While multivariate decoding algorithms are very useful for prediction their utility for interpretation may be limited when predictors are correlated.

Random walk of motor planning in task-irrelevant dimensions

The movements that we make are variable. It is well established that at least a part of this variability is caused by noise in central motor planning. Here, we studied how the random effects of planning noise translate into changes in motor planning. Are the random effects independently added to a constant mean end point, or do they accumulate over movements? To distinguish between these possibilities, we examined repeated, discrete movements in various tasks in which the motor output could be decomposed into a task-relevant and a task-irrelevant component. We found in all tasks that the task-irrelevant component had a positive lag 1 autocorrelation, suggesting that the random effects of planning noise accumulate over movements. In contrast, the task-relevant component always had a lag 1 autocorrelation close to zero, which can be explained by effective trial-by-trial correction of motor planning on the basis of observed motor errors. Accumulation of the effects of planning noise is consistent with current insights into the stochastic nature of synaptic plasticity. It leads to motor exploration, which may subserve motor learning and performance optimization.

Error probability of intracranial brain computer interfaces under non-task elicited brain states

OBJECTIVE

Intracranial brain computer interfaces (BCIs) can be connected to the user's cortex permanently. The interfaces response when fed with non-task elicited brain activity becomes important as design criterion: ideally intracranial BCIs should remain silent. We study their error probability in the form of false alarms.

METHODS

Using electrocorticograms recorded during task and non-task brain states, we compute false alarms, investigate their origin and introduce strategies to reduce them, using signal detection theory, classifier cascading and adaptation concepts.

RESULTS

We show that the incessant dynamics of the brain is prone to spontaneously produce signals, the spectral and topographical characteristics of which can resemble those associated with common control tasks, generating brain state classification errors.

CONCLUSIONS

In addition to hit and bit rates, response of BCIs to non-task brain states constitutes an important measure of BCI performance. Static classification cascading reduces considerably false positives during no-task brain states.

SIGNIFICANCE

False alarms in intracranial BCIs are undesirable and could have dangerous consequences for the users. Designs which effectively incorporate the error correction strategies discussed in this paper, could be more successful when taken from the laboratory or acute care setting and used in the real world.