Showing papers by "Fabien Lotte published in 2011"

Regularizing Common Spatial Patterns to Improve BCI Designs: Unified Theory and New Algorithms

Abstract: One of the most popular feature extraction algorithms for brain-computer interfaces (BCI) is common spatial patterns (CSPs). Despite its known efficiency and widespread use, CSP is also known to be very sensitive to noise and prone to overfitting. To address this issue, it has been recently proposed to regularize CSP. In this paper, we present a simple and unifying theoretical framework to design such a regularized CSP (RCSP). We then present a review of existing RCSP algorithms and describe how to cast them in this framework. We also propose four new RCSP algorithms. Finally, we compare the performances of 11 different RCSP (including the four new ones and the original CSP), on electroencephalography data from 17 subjects, from BCI competition datasets. Results showed that the best RCSP methods can outperform CSP by nearly 10% in median classification accuracy and lead to more neurophysiologically relevant spatial filters. They also enable us to perform efficient subject-to-subject transfer. Overall, the best RCSP algorithms were CSP with Tikhonov regularization and weighted Tikhonov regularization, both proposed in this paper.

Comparative study of band-power extraction techniques for Motor Imagery classification

Abstract: We review different techniques for extracting the power information contained in frequency bands in the context of electroencephalography (EEG) based Brain-Computer Interfaces (BCI). In this domain it is common to apply only one algorithm for extracting the power information. However previous work and our current study confirm that one may indeed expect varying degrees of success by choosing inadequate algorithms for the power extraction. Our results suggest that on average one algorithm seems superior for extracting the power information for Motor Imagery tasks : the application of a Morlet wavelet on the raw EEG signals, with the time-frequency resolution tradeoff selected by cross-validation.

Brain-computer interfaces for 3D games: hype or hope?

Abstract: Brain-Computer Interfaces (BCI) are communication systems conveying messages through brain activity only. This paper elaborates on the suitability of BCI for 3D Video Games (VG). Thus, we first review some recent BCI-based 3D VG. We then discuss the limitations of current BCI technology, those being mainly related to usability and performances. Finally, we report on some areas in which BCI could be useful for 3D VG despite their limitations. More precisely, BCI could be useful as an additional control channel, to send commands that cannot be intuitively sent with other devices. BCI could also be used for mental state monitoring either 1) during the game, in order to make adaptive and dynamic video games or 2) during the game creation in order to maximizes some measures of game quality that could be derived from a tester's mental state.

Generating Artificial EEG Signals To Reduce BCI Calibration Time

Abstract: One of the major limitations of Brain-Computer Interfaces (BCI) is their long calibration time. This is due to the need to collect numerous training EEG trials for the machine learning algorithm used in their design. In this paper we propose a new approach to reduce this calibration time. This approach consists in generating arti ficial EEG trials from the few EEG trials initially available, in order to augment the training set size in a relevant way. The approach followed is simple and computationally efficient. Moreover, our offline evaluations suggested that it can lead to signi ficant increases in classification accuracy when compared with existing approaches, especially when the number of training trials available is small. As such, it can indeed be used to reduce calibration time.

Using scalp electrical biosignals to control an object by concentration and relaxation tasks: Design and evaluation

Abstract: In this paper we explore the use of electrical biosignals measured on scalp and corresponding to mental relaxation and concentration tasks in order to control an object in a video game. To evaluate the requirements of such a system in terms of sensors and signal processing we compare two designs. The first one uses only one scalp electroencephalographic (EEG) electrode and the power in the alpha frequency band. The second one uses sixteen scalp EEG electrodes and machine-learning methods. The role of muscular activity is also evaluated using five electrodes positioned on the face and the neck. Results show that the first design enabled 70% of the participants to successfully control the game, whereas 100% of the participants managed to do it with the second design based on machine learning. Subjective questionnaires confirm these results: users globally felt to have control in both designs, with an increased feeling of control in the second one. Offline analysis of face and neck muscle activity shows that this activity could also be used to distinguish between relaxation and concentration tasks. Results suggest that the combination of muscular and brain activity could improve performance of this kind of system. They also suggest that muscular activity has probably been recorded by EEG electrodes.