We summarize results from a series of related studies that aim to develop a motor-imagery-based brain-computer interface using a single recording session of EEG or ECoG signals for each subject. We apply the same experimental and analytical methods to 11 non-paralysed subjects (8 EEG, 3 ECoG), and to 5 paralysed subjects (4 EEG, 1 ECoG) who had been unable to communicate for some time. While it was relatively easy to obtain classifiable signals quickly from most of the non-paralysed subjects, it proved impossible to classify the signals obtained from the paralysed patients by the same methods. This highlights the fact that though certain BCI paradigms may work well with healthy subjects, this does not necessarily indicate success with the target user group. We outline possible reasons for this failure to transfer.

project develops a non-invasive BCI system whose key features are (1) the use of well-established motor competences as control paradigms, (2) high-dimensional features from multi-channel EEG and (3) advanced machine learning techniques. Spatiospectral changes of sensorimotor rhythms are used to discriminate imagined movements (left hand, right hand, foot). A previous feedback study ([1]) with 10 subjects provided preliminary evidence that the BBCI system can be operated at high accuracy for subjects with less than 5 prior BCI exposures. Here, we demonstrate in a group of 14 fully BCI-naïve subjects that 8/14 BCI novices can perform at>84 % accuracy in their very first BCI session, and a further 4 subjects>70%. Thus, 12/14 BCI-novices had significant above-chance level performances without any subject training even in the first session, as based on an optimized EEG analysis by advanced machine learning algorithms. I.

In this paper we investigate the classification of mental tasks based on electroencephalographic (EEG) data for Brain Computer Interfaces (BCI) in two scenarios: off line and on-line. In the off-line scenario we evaluate the performance of a number of classifiers using a benchmark dataset, the same pre-processing and feature selection and show that classifiers that haven’t been used before are good choices. We also apply a new feature selection method that is suitable for the highly correlated EEG data and show that it greatly reduces the number of features without deteriorating the classification accuracy. In the on-line scenario that we have designed, we study the performance of our system to play a computer game for which the signals are processed in real time and the subject receives visual feedback of the resulting control within the game environment. We discuss the performance and highlight important issues.

Classifying electroencephalography (EEG) signals is an important step for proceeding EEG-based brain computer interfaces (BCI). Currently, kernel based methods such as support vector machine (SVM) are the state-of-the-art methods for this problem. In this paper, we apply Gaussian process (GP) classification to binary classification problems of motor imagery EEG data. Comparing with SVM, GP based methods naturally provide probability outputs for identifying a trusted prediction which can be used for post-processing in a BCI. Experimental results show that the classification methods based on Gaussian process perform similar with kernel logistic regression and probabilistic SVM in terms of predictive likelihood, but outperform SVM and K-Nearest Neighbor (KNN) in terms of 0-1 loss class prediction error.

Abstract. In this paper we empirically evaluate feature selection methods for classification of Brain-Computer Interface (BCI) data. We selected five state-of the-art methods, suitable for the noisy, correlated and highly dimensional BCI data, namely: information gain ranking, correlation-based feature selection, ReliefF, consistency-based feature selection and 1R ranking. We tested them with ten classification algorithms, representing different learning paradigms, on a benchmark BCI competition dataset. The results show that all feature selectors significantly reduced the number of features and also improved accuracy when used with suitable classification algorithms. The top three feature selectors in terms of classification accuracy were correlation-based feature selection, information gain and 1R ranking, with correlation based feature selection choosing the smallest number of features.

Abstract. In this paper we review classification algorithms used to design Brain-Computer Interface (BCI) systems based on ElectroEncephaloGraphy (EEG). We briefly present the commonly employed algorithms and describe their critical properties. Based on the literature, we compare them in terms of performance and provide guidelines to choose the suitable classification algorithm(s) for a specific BCI. PACS numbers: 8435, 8780 1.

Abstract—Brain-Computer Interfaces (BCIs) allow users to control applications by brain activity. Among their possible applications for non-disabled people, games are promising candidates. BCIs can enrich game play by the mental and affective state information they contain. During the eNTERFACE’09 workshop we developed the Bacteria Hunt game which can be played by keyboard and BCI, using SSVEP and relative alpha power. We conducted experiments in order to investigate what difference positive vs. negative neurofeedback would have on subjects ’ relaxation states and how well the different BCI paradigms can be used together. We observed no significant difference in mean alpha band power, thus relaxation, and in user experience between the games applying positive and negative feedback. We also found that alpha power before SSVEP stimulation was significantly higher than alpha power during SSVEP stimulation indicating that there is some interference between the two BCI paradigms. Index Terms—brain-computer interfaces, computer games, multimodal interaction.

acceptée sur proposition du jury:

Abstract. Electroencephalographic Evoked Response Potentials (ERP)s exhibit distinct and individualized spatial and temporal characteristics. Identification of spatio-temporal features improves single-trial classification performance and allows a better understanding of the underlying physiology. Thispaperpresentsamethodforanalyzingthespatio-temporal characteristics associated with Error related Potentials (ErrP)s. First, a resampling procedure based on Global Field Power (GFP) extracts temporal features. Second, a spatially weighted SVM (sw-SVM) is proposed to learn a spatial filter optimizing the classification performance for each temporal feature. Third, the so obtained ensemble of sw-SVM classifiers are combined using a weighted combination of all sw-SVM outputs. Results indicate that inclusion of temporal features provides useful insight regarding the spatio-temporal characteristics of error potentials. 1