BCI Competition 2003—Data Set IIb: Support Vector Machines for the P300 Speller Paradigm

A method for optimizing EEG electrode number and configuration for signal acquisition in P300 speller systems

OBJECTIVE

The P300 speller is intended to restore communication to patients with advanced neuromuscular disorders, but clinical implementation may be hindered by several factors, including system setup, burden, and cost. Our goal was to develop a method that can overcome these barriers by optimizing EEG electrode number and placement for P300 studies within a population of subjects.

METHODS

A Gibbs sampling method was developed to find the optimal electrode configuration given a set of P300 speller data. The method was tested on a set of data from 15 healthy subjects using an established 32-electrode pattern. Resulting electrode configurations were then validated using online prospective testing with a naïve Bayes classifier in 15 additional healthy subjects.

RESULTS

The method yielded a set of four posterior electrodes (PO₈, PO₇, POZ, CPZ), which produced results that are likely sufficient to be clinically effective. In online prospective validation testing, no significant difference was found between subjects' performances using the reduced and the full electrode configurations.

CONCLUSIONS

The proposed method can find reduced sets of electrodes within a subject population without reducing performance.

SIGNIFICANCE

Reducing the number of channels may reduce costs, set-up time, signal bandwidth, and computation requirements for practical online P300 speller implementation.

Preprocessing by a Bayesian single-trial event-related potential estimation technique allows feasibility of an assistive single-channel P300-based brain-computer interface

A major clinical goal of brain-computer interfaces (BCIs) is to allow severely paralyzed patients to communicate their needs and thoughts during their everyday lives. Among others, P300-based BCIs, which resort to EEG measurements, have been successfully operated by people with severe neuromuscular disabilities. Besides reducing the number of stimuli repetitions needed to detect the P300, a current challenge in P300-based BCI research is the simplification of system's setup and maintenance by lowering the number N of recording channels. By using offline data collected in 30 subjects (21 amyotrophic lateral sclerosis patients and 9 controls) through a clinical BCI with N = 5 channels, in the present paper we show that a preprocessing approach based on a Bayesian single-trial ERP estimation technique allows reducing N to 1 without affecting the system's accuracy. The potentially great benefit for the practical usability of BCI devices (including patient acceptance) that would be given by the reduction of the number N of channels encourages further development of the present study, for example, in an online setting.

Motor Imagery EEG Recognition Based on WPD-CSP and KF-SVM in Brain Computer Interfaces

A recognition method based on Wavelet Packet Decomposition - Common Spatial Patterns (WPD-CSP) and Kernel Fisher Support Vector Machine (KF-SVM) is developed and used for EEG recognition in motor imagery brain–computer interfaces (BCIs). The WPD-CSP is used for feature extraction and KF-SVM is used for classification. The presented recognition method includes the following steps: (1) some important EEG channels are selected. The 'haar' wavelet basis is used to take wavelet packet decomposition. And some decomposed sub-bands related with motor imagery for each EEG channel are reconstructed to obtain the relevant frequency information. (2) A six-dimensional feature vector is obtained by the CSP feature extraction to the reconstructed signal. And then the within-class scatter is calculated based on the feature vector. (3) The scatter is added into the radical basis function to construct a new kernel function. The obtained new kernel is integrated into the SVM to act as its kernel function. To evaluate effectiveness of the proposed WPD-CSP + KF-SVM method, the data from the 2008 international BCI competition are processed. A preliminary result shows that the proposed classification algorithm can well recognize EEG data and improve the EEG recognition accuracy in motor imagery BCIs.

An automated and fast approach to detect single-trial visual evoked potentials with application to brain-computer interface

OBJECTIVE

This study aims (1) to develop an automated and fast approach for detecting visual evoked potentials (VEPs) in single trials and (2) to apply the single-trial VEP detection approach in designing a real-time and high-performance brain-computer interface (BCI) system.

METHODS

The single-trial VEP detection approach uses common spatial pattern (CSP) as a spatial filter and wavelet filtering (WF) a temporal-spectral filter to jointly enhance the signal-to-noise ratio (SNR) of single-trial VEPs. The performance of the joint spatial-temporal-spectral filtering approach was assessed in a four-command VEP-based BCI system.

RESULTS

The offline classification accuracy of the BCI system was significantly improved from 67.6±12.5% (raw data) to 97.3±2.1% (data filtered by CSP and WF). The proposed approach was successfully implemented in an online BCI system, where subjects could make 20 decisions in one minute with classification accuracy of 90%.

CONCLUSIONS

The proposed single-trial detection approach is able to obtain robust and reliable VEP waveform in an automatic and fast way and it is applicable in VEP based online BCI systems.

SIGNIFICANCE

This approach provides a real-time and automated solution for single-trial detection of evoked potentials or event-related potentials (EPs/ERPs) in various paradigms, which could benefit many applications such as BCI and intraoperative monitoring.

Electrode subset selection methods for an EEG-based P300 brain-computer interface

PURPOSE

An electroencephalography (EEG)-based P300 speller is a type of brain-computer interface (BCI) that uses EEG to allow a user to select characters without physical movement. In general, using fewer electrodes for such a system makes it easier to set up and less expensive. This study addresses the question of electrode selection for EEG-based P300 systems.

METHODS

Data from 13 subjects collected with a 16-electrode cap was analyzed. The optimal subsets of electrodes of sizes 1-15 were calculated for each subject and for the group as a whole. The methods of exhaustive search, forward selection, and backward elimination were then compared to each other and to these optimal subsets.

RESULTS

The results show that, while none of the methods consistently picked the best-performing electrode subsets, all methods were able to find small electrode subsets that provided acceptable accuracy both for individuals and for the whole group. The computationally intensive exhaustive search method provided no statistically significant increase in performance over the much quicker forward and backward selection methods.

CONCLUSIONS

The forward and backward selection methods are preferred for electrode selection.

IMPLICATIONS FOR REHABILITATION

A P300 speller is a type of brain-computer interface that allows a user to select characters without physical movement. Using fewer electrodes reduces setup time and cost for an EEG-based P300 speller. We show that acceptable P300 speller performance can be achieved with as few as four electrodes. We compare methods of selecting electrode sets and identify fast and efficient methods for customizing electrode sets for individuals.

Integrating language information with a hidden Markov model to improve communication rate in the P300 speller

The P300 speller is a common brain-computer interface (BCI) application designed to communicate language by detecting event related potentials in a subject's electroencephalogram (EEG) signal. Information about the structure of natural language can be valuable for BCI communication systems, but few attempts have been made to incorporate this domain knowledge into the classifier. In this study, we treat BCI communication as a hidden Markov model (HMM) where hidden states are target characters and the EEG signal is the visible output. Using the Viterbi algorithm, language information can be incorporated in classification and errors can be corrected automatically. This method was first evaluated offline on a dataset of 15 healthy subjects who had a significant increase in bit rate from a previously published naïve Bayes approach and an average 32% increase from standard classification with dynamic stopping. An online pilot study of five healthy subjects verified these results as the average bit rate achieved using the HMM method was significantly higher than that using the naïve Bayes and standard methods. These findings strongly support the integration of domain-specific knowledge into BCI classification to improve system performance and accuracy.

A P300-based brain computer interface system for words typing

P300 is an event related potential of the brain in response to oddball events. Brain Computer Interface (BCI) utilizing P300 is known as a P300 BCI system. A conventional P300 BCI system for character spelling is composed of a paradigm that displays flashing characters and a classification scheme which identifies target characters. To type a word a user has to spell each character of the word: this spelling process is slow and it can take several minutes to type a word. In this study, we propose a new word typing scheme by integrating a word suggestion mechanism with a dictionary search into the conventional P300-based speller. Our new P300-based word typing system consists of an initial character spelling paradigm, a dictionary unit to give suggestions of possible words and the second word selection paradigm to select a word out of the suggestions. Our proposed methodology reduces typing time significantly and makes word typing easy via a P300 BCI system. We have tested our system with ten subjects and our results demonstrate an average word typing time of 1.91 min whereas the conventional took 3.36 min for the same words.

Aggregation of sparse linear discriminant analyses for event-related potential classification in brain-computer interface

Two main issues for event-related potential (ERP) classification in brain-computer interface (BCI) application are curse-of-dimensionality and bias-variance tradeoff, which may deteriorate classification performance, especially with insufficient training samples resulted from limited calibration time. This study introduces an aggregation of sparse linear discriminant analyses (ASLDA) to overcome these problems. In the ASLDA, multiple sparse discriminant vectors are learned from differently l1-regularized least-squares regressions by exploiting the equivalence between LDA and least-squares regression, and are subsequently aggregated to form an ensemble classifier, which could not only implement automatic feature selection for dimensionality reduction to alleviate curse-of-dimensionality, but also decrease the variance to improve generalization capacity for new test samples. Extensive investigation and comparison are carried out among the ASLDA, the ordinary LDA and other competing ERP classification algorithms, based on different three ERP datasets. Experimental results indicate that the ASLDA yields better overall performance for single-trial ERP classification when insufficient training samples are available. This suggests the proposed ASLDA is promising for ERP classification in small sample size scenario to improve the practicability of BCI.

Evaluating true BCI communication rate through mutual information and language models

Brain-computer interface (BCI) systems are a promising means for restoring communication to patients suffering from “locked-in” syndrome. Research to improve system performance primarily focuses on means to overcome the low signal to noise ratio of electroencephalogric (EEG) recordings. However, the literature and methods are difficult to compare due to the array of evaluation metrics and assumptions underlying them, including that: 1) all characters are equally probable, 2) character selection is memoryless, and 3) errors occur completely at random. The standardization of evaluation metrics that more accurately reflect the amount of information contained in BCI language output is critical to make progress. We present a mutual information-based metric that incorporates prior information and a model of systematic errors. The parameters of a system used in one study were re-optimized, showing that the metric used in optimization significantly affects the parameter values chosen and the resulting system performance. The results of 11 BCI communication studies were then evaluated using different metrics, including those previously used in BCI literature and the newly advocated metric. Six studies' results varied based on the metric used for evaluation and the proposed metric produced results that differed from those originally published in two of the studies. Standardizing metrics to accurately reflect the rate of information transmission is critical to properly evaluate and compare BCI communication systems and advance the field in an unbiased manner.

Channel selection based on phase measurement in P300-based brain-computer interface

Most EEG-based brain-computer interface (BCI) paradigms include specific electrode positions. As the structures and activities of the brain vary with each individual, contributing channels should be chosen based on original records of BCIs. Phase measurement is an important approach in EEG analyses, but seldom used for channel selections. In this paper, the phase locking and concentrating value-based recursive feature elimination approach (PLCV-RFE) is proposed to produce robust-EEG channel selections in a P300 speller. The PLCV-RFE, deriving from the phase resetting mechanism, measures the phase relation between EEGs and ranks channels by the recursive strategy. Data recorded from 32 electrodes on 9 subjects are used to evaluate the proposed method. The results show that the PLCV-RFE substantially reduces channel sets and improves recognition accuracies significantly. Moreover, compared with other state-of-the-art feature selection methods (SSNRSF and SVM-RFE), the PLCV-RFE achieves better performance. Thus the phase measurement is available in the channel selection of BCI and it may be an evidence to indirectly support that phase resetting is at least one reason for ERP generations.

Bayesian approach to dynamically controlling data collection in P300 spellers

P300 spellers provide a noninvasive method of communication for people who may not be able to use other communication aids due to severe neuromuscular disabilities. However, P300 spellers rely on event-related potentials (ERPs) which often have low signal-to-noise ratios (SNRs). In order to improve detection of the ERPs, P300 spellers typically collect multiple measurements of the electroencephalography (EEG) response for each character. The amount of collected data can affect both the accuracy and the communication rate of the speller system. The goal of the present study was to develop an algorithm that would automatically determine the necessary amount of data to collect during operation. Dynamic data collection was controlled by a threshold on the probabilities that each possible character was the target character, and these probabilities were continually updated with each additional measurement. This Bayesian technique differs from other dynamic data collection techniques by relying on a participant-independent, probability-based metric as the stopping criterion. The accuracy and communication rate for dynamic and static data collection in P300 spellers were compared for 26 users. Dynamic data collection resulted in a significant increase in accuracy and communication rate.

Subliminal salience search illustrated: EEG identity and deception detection on the fringe of awareness

We propose a novel deception detection system based on Rapid Serial Visual Presentation (RSVP). One motivation for the new method is to present stimuli on the fringe of awareness, such that it is more difficult for deceivers to confound the deception test using countermeasures. The proposed system is able to detect identity deception (by using the first names of participants) with a 100% hit rate (at an alpha level of 0.05). To achieve this, we extended the classic Event-Related Potential (ERP) techniques (such as peak-to-peak) by applying Randomisation, a form of Monte Carlo resampling, which we used to detect deception at an individual level. In order to make the deployment of the system simple and rapid, we utilised data from three electrodes only: Fz, Cz and Pz. We then combined data from the three electrodes using Fisher's method so that each participant was assigned a single p-value, which represents the combined probability that a specific participant was being deceptive. We also present subliminal salience search as a general method to determine what participants find salient by detecting breakthrough into conscious awareness using EEG.

An efficient words typing P300-BCI system using a modified T9 interface and random forest classifier

The conventional P300-based character spelling BCI system consists of a character presentation paradigm and a classification system. In this paper, we propose modifications to both in order to increase the word typing speed and accuracy. In the paradigm part, we have modified the T9 (Text on Nine keys) interface which is similar to the keypad of mobile phones being used for text messaging. Then we have integrated a custom-built dictionary to give word suggestions to a user while typing. The user can select one out of the given suggestions to complete word typing. Our proposed paradigms significantly reduce the word typing time and make words typing more convenient by typing complete words with only few initial character spellings. In the classification part we have adopted a Random Forest (RF) classifier. The RF improves classification accuracy by combining multiple decision trees. We conducted experiments with five subjects using the proposed BCI system. Our results demonstrate that our system increases typing speed significantly: our proposed system took an average time of 1.83 minutes per word, while typing ten random words, whereas the conventional spelling required 3.35 minutes for the same words under the same conditions, decreasing the typing time by 45.37%.

Noise enhanced array signal detection in P300 speller paradigm using ICA-based subspace projections

This paper explores how noise can improve prediction accuracy of the Event-Related Potential (ERP) based on P300 signals. We propose an array of ICA-Based P300 processing systems with additive white Gaussian noise. The array system attains maximum accuracy when noise intensity is not zero and thus the system shows the stochastic resonance effect. The prediction accuracy increases as the number of stages of the array increases. Experimental results show that increasing the array size with the proper amount of noise can improve the accuracy of the original P300 signal detection using ICA-based subspace projection technique.

P300 brain-computer interface design for communication and control applications

This paper introduces the design of a P300-based brain-computer interface (BCI) system. Based on this system, two applications are implemented: a word speller and a remote control device, which are to assist physically disabled people to communicate and control. A number of specific implementation techniques are proposed to achieve good performance in terms of accuracy and reliability. The word speller can achieve a spelling rate of up to 4-6 letters per minute, while both applications achieve 99% accuracy in our experiments with healthy subjects.

A New Single Trial P300 Classification Method

P300-Speller is one of the most practical and widely used Brain Computer Interface (BCI) for locked-in people who are not able to communicate with others via traditional communication methods. Many signal processing techniques have been utilized in P300-Speller to restore the communication ability of these locked-in people. These techniques are capable of achieving high classification accuracy. However the classification accuracy dramatically decreases for single trial analysis. The reason for that is that the noises existing in the recorded signals are usually removed by averaging several trials. When only a single trial is available, averaging is no longer an option for de-noising. The “averaging” step becomes the bottle neck of P300 response detection which highly limits the processing speed. Researchers are looking for techniques that can accomplish the classification task in a single trial. In this work, a new, effective but simple processing technique for single trial electroencephalography (EEG) classification using variance analysis based method is presented. This method achieved an overall accuracy of 84.8% for single trial P300 response identification. When compared with a single trial stepwise linear discriminant analysis (SWLDA), the authors’ method in terms of overall accuracy is more accurate and the data communication speed is significantly improved.

A new feature extraction method for signal classification applied to cord dorsum potential detection

In the spinal cord of the anesthetized cat, spontaneous cord dorsum potentials (CDPs) appear synchronously along the lumbo-sacral segments. These CDPs have different shapes and magnitudes. Previous work has indicated that some CDPs appear to be specially associated with the activation of spinal pathways that lead to primary afferent depolarization and presynaptic inhibition. Visual detection and classification of these CDPs provides relevant information on the functional organization of the neural networks involved in the control of sensory information and allows the characterization of the changes produced by acute nerve and spinal lesions. We now present a novel feature extraction approach for signal classification, applied to CDP detection. The method is based on an intuitive procedure. We first remove by convolution the noise from the CDPs recorded in each given spinal segment. Then, we assign a coefficient for each main local maximum of the signal using its amplitude and distance to the most important maximum of the signal. These coefficients will be the input for the subsequent classification algorithm. In particular, we employ gradient boosting classification trees. This combination of approaches allows a faster and more accurate discrimination of CDPs than is obtained by other methods.

The effects of stimulus timing features on P300 speller performance

OBJECTIVE

Despite numerous examinations of factors affecting P300 speller performance, the impact of stimulus presentation parameters remains incompletely understood. This study examines the effects of four distinct stimulus presentation parameters (stimulus-off time [ISI(∗)], interstimulus interval [ISI], flash duration, and flash-duration:ISI ratio) on the accuracy and efficiency of the P300 speller performance.

METHODS

EEG data from a 32-electrode set were recorded from six subjects using a row-column paradigm of the speller task and analyzed offline.

RESULTS

P300 speller accuracy is affected by the number of trial repetitions (F(14,354) = 69.002, p < 0.0001), as expected. In addition, longer ISI and ISI(∗) times resulted in higher accuracy and characters per minute [CPM] rates. Subsets of the entire group (i.e. good vs. poor performers) were compared to show consistency of performance trends despite great variance among subjects. Moreover, the same significant effects were observed whether using the entire 32-electrode dataset or the reduced 8-channel set described by Sharbrough et al. (1991).

CONCLUSIONS

Despite variability in user performance, both ISI(∗) and ISI can affect P300 speller performance.

SIGNIFICANCE

P300 system optimization must consider critical stimulus timing features including ISI(∗) and ISI. Further characterization of the impact of these timing features in online experiments is warranted and the differential effect on accuracy and CPM should be more comprehensively explored.

Review of the BCI Competition IV

The BCI competition IV stands in the tradition of prior BCI competitions that aim to provide high quality neuroscientific data for open access to the scientific community. As experienced already in prior competitions not only scientists from the narrow field of BCI compete, but scholars with a broad variety of backgrounds and nationalities. They include high specialists as well as students. The goals of all BCI competitions have always been to challenge with respect to novel paradigms and complex data. We report on the following challenges: (1) asynchronous data, (2) synthetic, (3) multi-class continuous data, (4) session-to-session transfer, (5) directionally modulated MEG, (6) finger movements recorded by ECoG. As after past competitions, our hope is that winning entries may enhance the analysis methods of future BCIs.

Classification of multichannel EEG patterns using parallel hidden Markov models

In this paper, a parallel hidden-Markov-model (PHMM)-based approach is proposed for the problem of multichannel electroencephalogram (EEG) patterns classification. The approach is based on multi-channel representation of the EEG signals using a parallel combination of HMMs, where each model represents a particular channel. The performance of the proposed algorithm is studied using an artificial EEG database, and two real EEG databases: a database of two classes of EEGs elicited during a task of imagery of hand upward and downward movements of a computer screen cursor (db Ia), and a database of two classes of sensorimotor EEGs elicited during a feedback-regulated left-right motor imagery task (db III). The results show that the proposed algorithm outperforms other commonly used methods with classification rate improvement of 2 and 10% for db Ia and db III, respectively. In addition, the proposed method outperforms a support vector machine classifier with a linear kernel, when both classifiers utilize the same feature set. The results also show that a model architecture which includes a left-to-right scheme with no skips, five states and three Gaussians, outperforms the other tested architectures due to the fact that it allows a better modeling of the temporal sequencing of the EEG components.

Brain computer interfaces, a review

A brain-computer interface (BCI) is a hardware and software communications system that permits cerebral activity alone to control computers or external devices. The immediate goal of BCI research is to provide communications capabilities to severely disabled people who are totally paralyzed or 'locked in' by neurological neuromuscular disorders, such as amyotrophic lateral sclerosis, brain stem stroke, or spinal cord injury. Here, we review the state-of-the-art of BCIs, looking at the different steps that form a standard BCI: signal acquisition, preprocessing or signal enhancement, feature extraction, classification and the control interface. We discuss their advantages, drawbacks, and latest advances, and we survey the numerous technologies reported in the scientific literature to design each step of a BCI. First, the review examines the neuroimaging modalities used in the signal acquisition step, each of which monitors a different functional brain activity such as electrical, magnetic or metabolic activity. Second, the review discusses different electrophysiological control signals that determine user intentions, which can be detected in brain activity. Third, the review includes some techniques used in the signal enhancement step to deal with the artifacts in the control signals and improve the performance. Fourth, the review studies some mathematic algorithms used in the feature extraction and classification steps which translate the information in the control signals into commands that operate a computer or other device. Finally, the review provides an overview of various BCI applications that control a range of devices.

Input interface using event-related potential P3

This paper refers to a basic study toward the goal of developing a simple and easy-to-use input interface based on P3 components of visual, event-related potentials. Because contamination from eye movements and eye blinks is a problem, a method for removing eye movement artifacts from electroencephalogram (EEG) signals by applying an independent component analysis un-mixing matrix was proposed and implemented. Input character decisions were executed using a support vector machine (SVM) for judging the P3 existence of a single stimulus. The performances were compared while varying the number of channels of EEG signals, the types of feature vectors, and the ratio of the number of data used for training the SVM. The results indicated that three EEG signal channels (Fz, Cz, Pz) were enough to remove artifacts related to eye blinks and vertical eye movements and could be used to make a decision about input characters. The number of trials necessary to decide the input characters was ten on average. The best ratio achieved for the number of training data of targets and non-targets was 1∶2. These results should be confirmed using a larger number of data sets.

Pattern Discovery from Biological Data

Extracting useful information from structured and unstructured biological data is crucial in the health industry. Some examples include medical practitioner’s need to identify breast cancer patient in the early stage, estimate survival time of a heart disease patient, or recognize uncommon disease characteristics which suddenly appear. Currently there is an explosion in biological data available in the data bases. But information extraction and true open access to data are require time to resolve issues such as ethical clearance. The emergence of novel IT technologies allows health practitioners to facilitate the comprehensive analyses of medical images, genomes, transcriptomes, and proteomes in health and disease. The information that is extracted from such technologies may soon exert a dramatic change in the pace of medical research and impact considerably on the care of patients. The current research will review the existing technologies being used in heart and cancer research. Finally this research will provide some possible solutions to overcome the limitations of existing technologies. In summary the primary objective of this research is to investigate how existing modern machine learning techniques (with their strength and limitations) are being used in the indent of heartbeat related disease and the early detection of cancer in patients. After an extensive literature review these are the objectives chosen: to develop a new approach to find the association between diseases such as high blood pressure, stroke and heartbeat, to propose an improved feature selection method to analyze huge images and microarray databases for machine learning algorithms in cancer research, to find an automatic distance function selection method for clustering tasks, to discover the most significant risk factors for specific cancers, and to determine the preventive factors for specific cancers that are aligned with the most significant risk factors. Therefore we propose a research plan to attain these objectives within this chapter. The possible solutions of the above objectives are: new heartbeat identification techniques show promising association with the heartbeat patterns and diseases, sensitivity based feature selection methods will be applied to early cancer patient classification, meta learning approaches will be adopted in clustering algorithms to select an automatic distance function, and Apriori algorithm will be applied to discover the significant risks and preventive factors for specific cancers. We expect this research will add significant contributions to the medical professional to enable more accurate diagnosis and better patient care. It will also contribute in other area such as biomedical modeling, medical image analysis and early diseases warning.

Natural language processing with dynamic classification improves P300 speller accuracy and bit rate

The P300 speller is an example of a brain-computer interface that can restore functionality to victims of neuromuscular disorders. Although the most common application of this system has been communicating language, the properties and constraints of the linguistic domain have not to date been exploited when decoding brain signals that pertain to language. We hypothesized that combining the standard stepwise linear discriminant analysis with a Naive Bayes classifier and a trigram language model would increase the speed and accuracy of typing with the P300 speller. With integration of natural language processing, we observed significant improvements in accuracy and 40-60% increases in bit rate for all six subjects in a pilot study. This study suggests that integrating information about the linguistic domain can significantly improve signal classification.

Evaluation of the performances of different P300 based brain-computer interfaces by means of the efficiency metric

The aim of this paper is to show how to use the Efficiency, a brain-computer interface (BCI) performance indicator, to evaluate the performances of a wide range of BCI systems. Unlike the most used metrics in the BCI research field, the Efficiency takes into account the penalties and the strategies to recover errors and this makes it a reliable instrument to describe the behavior of real BCIs. The Efficiency is compared with the accuracy and the information transfer rate, both in the Wolpaw and Nykopp definitions. The comparison covers four widely used classifiers and different stimulation sequences. Results show that the Efficiency is able to predict if the communication will not be possible, because the time spent to correct mistakes is longer than the time needed to generate a correct selection, and therefore it provides a much more realistic evaluation of a system. It can also be easily adapted to evaluate different applications, so it reveals a more general and versatile indicator for BCI systems.

Enabling fast brain-computer interaction by single-trial extraction of visual evoked potentials

This paper investigates the challenging issue of enabling fast brain-computer interaction to construct a mental speller. Exploiting visual evoked potentials as communication carriers, an online paradigm called "imitating-human-natural-reading" is realized. In this online paradigm, single-trial estimation with the intrinsically real-time feature should be used instead of grand average that is traditionally used in the cognitive or clinical experiments. By the use of several montages of component features from four channels with parameter optimization, we explored the support vector machines-based single-trial estimation of evoked potentials. The results on a human-subject show the advantages of the inducing paradigm used in our mental speller with a high classification rate.

Suppressing flashes of items surrounding targets during calibration of a P300-based brain-computer interface improves performance

Since the introduction of the P300 brain-computer interface (BCI) speller by Farwell and Donchin in 1988, the speed and accuracy of the system has been significantly improved. Larger electrode montages and various signal processing techniques are responsible for most of the improvement in performance. New presentation paradigms have also led to improvements in bit rate and accuracy (e.g. Townsend et al (2010 Clin. Neurophysiol. 121 1109-20)). In particular, the checkerboard paradigm for online P300 BCI-based spelling performs well, has started to document what makes for a successful paradigm, and is a good platform for further experimentation. The current paper further examines the checkerboard paradigm by suppressing items which surround the target from flashing during calibration (i.e. the suppression condition). In the online feedback mode the standard checkerboard paradigm is used with a stepwise linear discriminant classifier derived from the suppression condition and one classifier derived from the standard checkerboard condition, counter-balanced. The results of this research demonstrate that using suppression during calibration produces significantly more character selections/min ((6.46) time between selections included) than the standard checkerboard condition (5.55), and significantly fewer target flashes are needed per selection in the SUP condition (5.28) as compared to the RCP condition (6.17). Moreover, accuracy in the SUP and RCP conditions remained equivalent (∼90%). Mean theoretical bit rate was 53.62 bits/min in the suppression condition and 46.36 bits/min in the standard checkerboard condition (ns). Waveform morphology also showed significant differences in amplitude and latency.

An asynchronous P300 BCI with SSVEP-based control state detection

In this paper, an asynchronous brain-computer interface (BCI) system combining the P300 and steady-state visually evoked potentials (SSVEPs) paradigms is proposed. The information transfer is accomplished using P300 event-related potential paradigm and the control state (CS) detection is achieved using SSVEP, overlaid on the P300 base system. Offline and online experiments have been performed with ten subjects to validate the proposed system. It is shown to achieve fast and accurate CS detection without significantly compromising the performance. In online experiments, the system is found to be capable of achieving an average data transfer rate of 19.05 bits/min, with CS detection accuracy of about 88%.

Toward a model-based predictive controller design in brain-computer interfaces

A first step in designing a robust and optimal model-based predictive controller (MPC) for brain-computer interface (BCI) applications is presented in this article. An MPC has the potential to achieve improved BCI performance compared to the performance achieved by current ad hoc, nonmodel-based filter applications. The parameters in designing the controller were extracted as model-based features from motor imagery task-related human scalp electroencephalography. Although the parameters can be generated from any model-linear or non-linear, we here adopted a simple autoregressive model that has well-established applications in BCI task discriminations. It was shown that the parameters generated for the controller design can as well be used for motor imagery task discriminations with performance (with 8-23% task discrimination errors) comparable to the discrimination performance of the commonly used features such as frequency specific band powers and the AR model parameters directly used. An optimal MPC has significant implications for high performance BCI applications.

Predictive spelling with a P300-based brain-computer interface: Increasing the rate of communication

This study compared a conventional P300 speller brain-computer interface (BCI) to one used in conjunction with a predictive spelling program. Performance differences in accuracy, bit rate, selections per minute, and output characters per minute (OCM) were examined. An 8×9 matrix of letters, numbers, and other keyboard commands was used. Participants (n = 24) were required to correctly complete the same 58 character sentence (i.e., correcting for errors) using the predictive speller (PS) and the non-predictive speller (NS), counterbalanced. The PS produced significantly higher OCMs than the NS. Time to complete the task in the PS condition was 12min 43sec as compared to 20min 20sec in the NS condition. Despite the marked improvement in overall output, accuracy was significantly higher in the NS paradigm. P300 amplitudes were significantly larger in the NS than in the PS paradigm; which is attributed to increased workload and task demands. These results demonstrate the potential efficacy of predictive spelling in the context of BCI.

Out of the frying pan into the fire--the P300-based BCI faces real-world challenges

Brain-computer interfaces (BCIs) have been investigated for more than 20 years. Many BCIs use noninvasive electroencephalography as a measurement technique and the P300 event-related potential as an input signal (P300 BCI). Since the first experiment with a P300 BCI system in 1988 by Farwell and Donchin, not only data processing has improved but also stimuli presentation has been varied and a plethora of applications was developed and refined. Nowadays, these applications are facing the challenge of being transferred from the research laboratory into real-life situations to serve motor-impaired people in their homes as assistive technology.

Statistical spatial filtering for a P300-based BCI: tests in able-bodied, and patients with cerebral palsy and amyotrophic lateral sclerosis

The effective use of brain-computer interfaces (BCIs) in real-world environments depends on a satisfactory throughput. In a P300-based BCI, this can be attained by reducing the number of trials needed to detect the P300 signal. However, this task is hampered by the very low signal-to-noise-ratio (SNR) of P300 event related potentials. This paper proposes an efficient methodology that achieves high classification accuracy and high transfer rates for both disabled and able-bodied subjects in a standard P300-based speller system. The system was tested by three subjects with cerebral palsy (CP), two subjects with amyotrophic lateral sclerosis (ALS), and nineteen able-bodied subjects. The paper proposes the application of three statistical spatial filters. The first is a beamformer that maximizes the ratio of signal power and noise power (Max-SNR). The second is a beamformer based on the Fisher criterion (FC). The third approach cascades the FC beamformer with the Max-SNR beamformer satisfying simultaneously sub-optimally both criteria (C-FMS). The calibration process of the BCI system takes about 5 min to collect data and a couple of minutes to obtain spatial filters and classification models. Online results showed that subjects with disabilities have achieved, on average, an accuracy and transfer rate only slightly lower than able-bodied subjects. Taking 23 of the 24 participants, the averaged results achieved a transfer rate of 4.33 symbols per minute with a 91.80% accuracy, corresponding to a bandwidth of 19.18 bits per minute. This study shows the feasibility of the proposed methodology and that effective communication rates are achievable.

A graphical model framework for decoding in the visual ERP-based BCI speller

We present a graphical model framework for decoding in the visual ERP-based speller system. The proposed framework allows researchers to build generative models from which the decoding rules are obtained in a straightforward manner. We suggest two models for generating brain signals conditioned on the stimulus events. Both models incorporate letter frequency information but assume different dependencies between brain signals and stimulus events. For both models, we derive decoding rules and perform a discriminative training. We show on real visual speller data how decoding performance improves by incorporating letter frequency information and using a more realistic graphical model for the dependencies between the brain signals and the stimulus events. Furthermore, we discuss how the standard approach to decoding can be seen as a special case of the graphical model framework. The letter also gives more insight into the discriminative approach for decoding in the visual speller system.