Brain Computer Interfaces (BCIs) are systems developed in order to control devices by using only brain signals. In BCI systems, different mental activities to be performed by the users are associated with different actions on the device to be controlled. Spelling Paradigm is a BCI application which aims to construct the words by finding letters using P300 signals recorded via channel electrodes attached to the diverse points of the scalp. Reducing the letter detection error rates and increasing the speed of letter detection are crucial for Spelling Paradigm. By this way, disabled people can express their needs more easily using this application. In this thesis, two different methods, Support Vector Machine (SVM) and AdaBoost, are used for classification in the analysis. Classification and Regression Trees is used as the weak classifier of the AdaBoost. Time-frequency domain characteristics of P300 evoked potentials are analyzed in addition to time domain characteristics. Wigner-Ville Distribution is used for transforming time domain signals into time-frequency domain. It is observed that classification results are better in time domain. Furthermore, optimum subset of channels that models P300 signals with minimum error rate is searched. A method that uses both SVM and AdaBoost is proposed to select channels. 12 channels are selected in time domain with this method. Also, effect of dimension reduction is analyzed using Principal Component Analysis (PCA) and AdaBoost methods.

Les Interfaces Cerveau Machine (ICM) sont des dispositifs d'un type nouveau permettant la communication directe entre le cerveau d'un utilisateur et une machine. De tels dispositifs peuvent être réalisés grâce à la mesure non-invasive d'informations provenant du cortex par électroencéphalographie (EEG). Un des enjeux principaux du domaine est de réussir à extraire en temps réel, à partir d'une source d'information très limitée et bruitée, un signal de commande robuste et suffisamment complexe pour permettre le contrôle d'un programme ou d'un effecteur. Dans cette thèse nous avons contribué à l'amélioration des ICM sur trois points. Tout d'abord, nous avons exploré la possibilité d'augmenter la résolution spatiale de l'EEG en utilisant des méthodes permettant de construire l'activité corticale en temps réel. D'autre part, inspirés par une ICM utilisant l'imagination motrice des pieds pour envoyer une commande unique, nous nous sommes intéressés aux signaux produits par l'imagination de mouvements brefs. Cela nous a permis de développer un nouveau type d'ICM appelé bouton commandé par la pensée. Une telle ICM permet à l'utilisateur d'envoyer, de façon asynchrone, plusieurs commandes en imaginant différents mouvements. Finalement, nous avons développé une méthode, basée sur la théorie des bandits stochastiques, pour sélectionner automatiquement et efficacement le mouvement imaginé le plus discernable de l'état de repos et permettant ainsi le meilleur contrôle de l'ICM. En parallèle, nous avons développé une boîte à outils Matlab qui automatise l'ajustement des paramètres et la comparaison des différentes méthodes utilisées pour réaliser une ICM.

Προτείνουμε μια προσέγγιση για να αναλύσουμε τα δεδομένα που συλλέγουμε από το παράδειγμα του Ορθογράφου P300, χρησιμοποιώντας την τεχνική μηχανικής μάθησης, support vector machines. Στο συγκεκριμένο πλαίσιο κατηγοριοποίησης, πετύχαμε την σωστή λύση μετά από πέντε επαναλήψεις. Ενώ η κατηγοριοποίηση στους διαγωνισμούς BCI είναι για offline ανάλυση, η προσέγγιση μας μπορεί να θεωρηθεί ως μια online λύση για τον πραγματικό κόσμο. Είναι γρήγορη, απαιτεί λίγες (λιγότερες από 10) θέσεις ηλεκτροδίων, απαιτεί μόνο ένα μικρό ποσοστό προεπεξεργασίας και η επιλογή των τιμών για κρίσιμες παραμέτρους έχει αυτοματοποιηθεί. / We propose an approach to analyze data from the P300 speller paradigm using the machine-learning technique support vector machines. In a conservative classification scheme, we found the correct solution after five repetitions. While the classification within the BCI competition is designed for offline analysis, our approach is also well-suited for a real-world online solution: It is fast, requires few electrode positions (less than 10), demands only a small amount of preprocessing and selection of values for critical parameters is automated.

O uso de sistemas denominados Brain Computer Interface, ou simplesmente BCI, para controle de dispositivos tem gerado cada vez mais trabalhos de análise de sinais de EEG, principalmente devido ao fato do desenvolvimento tecnológico dos sistemas de processamento de dados, trazendo novas perspectiva de desenvolvimento de equipamentos que auxiliem pessoas com debilidades motoras. Neste trabalho é abordado o comportamento dos classificadores LDA (Discriminante Linear de Fisher) e o classificador Naive Bayes para classificação de movimento de mão direita e mão esquerda a partir da aquisição de sinais eletroencefalográficos. Para análise destes classificadores foram utilizadas como características de entrada a energia de trechos do sinal filtrados por um passa banda com frequências dentro dos ritmos sensório-motor e também foram utilizadas componentes de energia espectral através do periodograma modificado de Welch. Como forma de pré-processamento também é apresentado o filtro espacial Common Spatial Pattern (CSP) de forma a aumentar a atividade discriminativa entre as classes de movimento. Foram obtidas taxas de acerto de até 70% para a base de dados geradas neste trabalho e de até 88% utilizando a base de dados do BCI Competition II, taxas de acertos compatíveis com outros trabalhos na área. / Brain Computer Interface (BCI) systems usage for controlling devices has increasingly generated research on EEG signals analysis, mainly because the technological development of data processing systems has been offering a new perspective on developing equipment to assist people with motor disability. This study aims to examine the behavior of both Fisher's Linear Discriminant (LDA) and Naive Bayes classifiers in determining both the right and left hand movement through electroencephalographic signals. To accomplish this, we considered as input feature the energy of the signal trials filtered by a band pass with sensorimotor rhythm frequencies; spectral power components from the Welch modified periodogram were also used. As a preprocessing form, the Common Spatial Pattern (CSP) filter was used to increase the discriminative activity between classes of movement. The database created from this study reached hit rates of up to 70% while the BCI Competition II reached hit rates up to 88%, which is consistent with the literature.

Brain-computer interfaces (BCIs) blend methods and concepts researched by cognitive neuroscience, electrophysiology, computer science and engineering, resulting in systems of bi-directional information exchange directly between brain and computer. BCIs contribute to medical applications that restore communication and mobility for disabled patients and provide new forms of sending information to devices for enhancement and entertainment. Virtual reality (VR) introduces humans into a computer-generated world, tackling immersion and involvement. VR technology extends the classical multimedia experience, as the user is able to move within the environment, interact with other virtual participants, and manipulate objects, in order to generate the feeling of presence. This essay presents the possibilities of merging BCI with VR and the challenges to be tackled in the future. Current attempts to combine BCI and VR technology have shown that VR is a useful tool to test the functioning of BCIs, with safe, controlled and realistic experiments; there are better outcomes for VR and BCI combinations used for medical purposes compared to solely BCI training; and, enhancement systems for healthy users seem promising with VR-BCIs designed for home users. Future trends include brain-to-brain communication, sharing of several users’ brain signals within the virtual environment, and better and more efficient interfaces.

O uso de sistemas denominados Brain Computer Interface, ou simplesmente BCI, para controle de dispositivos tem gerado cada vez mais trabalhos de análise de sinais de EEG, principalmente devido ao fato do desenvolvimento tecnológico dos sistemas de processamento de dados, trazendo novas perspectiva de desenvolvimento de equipamentos que auxiliem pessoas com debilidades motoras. Neste trabalho é abordado o comportamento dos classificadores LDA (Discriminante Linear de Fisher) e o classificador Naive Bayes para classificação de movimento de mão direita e mão esquerda a partir da aquisição de sinais eletroencefalográficos. Para análise destes classificadores foram utilizadas como características de entrada a energia de trechos do sinal filtrados por um passa banda com frequências dentro dos ritmos sensório-motor e também foram utilizadas componentes de energia espectral através do periodograma modificado de Welch. Como forma de pré-processamento também é apresentado o filtro espacial Common Spatial Pattern (CSP) de forma a aumentar a atividade discriminativa entre as classes de movimento. Foram obtidas taxas de acerto de até 70% para a base de dados geradas neste trabalho e de até 88% utilizando a base de dados do BCI Competition II, taxas de acertos compatíveis com outros trabalhos na área. / Brain Computer Interface (BCI) systems usage for controlling devices has increasingly generated research on EEG signals analysis, mainly because the technological development of data processing systems has been offering a new perspective on developing equipment to assist people with motor disability. This study aims to examine the behavior of both Fisher's Linear Discriminant (LDA) and Naive Bayes classifiers in determining both the right and left hand movement through electroencephalographic signals. To accomplish this, we considered as input feature the energy of the signal trials filtered by a band pass with sensorimotor rhythm frequencies; spectral power components from the Welch modified periodogram were also used. As a preprocessing form, the Common Spatial Pattern (CSP) filter was used to increase the discriminative activity between classes of movement. The database created from this study reached hit rates of up to 70% while the BCI Competition II reached hit rates up to 88%, which is consistent with the literature.

O uso de sistemas denominados Brain Computer Interface, ou simplesmente BCI, para controle de dispositivos tem gerado cada vez mais trabalhos de análise de sinais de EEG, principalmente devido ao fato do desenvolvimento tecnológico dos sistemas de processamento de dados, trazendo novas perspectiva de desenvolvimento de equipamentos que auxiliem pessoas com debilidades motoras. Neste trabalho é abordado o comportamento dos classificadores LDA (Discriminante Linear de Fisher) e o classificador Naive Bayes para classificação de movimento de mão direita e mão esquerda a partir da aquisição de sinais eletroencefalográficos. Para análise destes classificadores foram utilizadas como características de entrada a energia de trechos do sinal filtrados por um passa banda com frequências dentro dos ritmos sensório-motor e também foram utilizadas componentes de energia espectral através do periodograma modificado de Welch. Como forma de pré-processamento também é apresentado o filtro espacial Common Spatial Pattern (CSP) de forma a aumentar a atividade discriminativa entre as classes de movimento. Foram obtidas taxas de acerto de até 70% para a base de dados geradas neste trabalho e de até 88% utilizando a base de dados do BCI Competition II, taxas de acertos compatíveis com outros trabalhos na área. / Brain Computer Interface (BCI) systems usage for controlling devices has increasingly generated research on EEG signals analysis, mainly because the technological development of data processing systems has been offering a new perspective on developing equipment to assist people with motor disability. This study aims to examine the behavior of both Fisher's Linear Discriminant (LDA) and Naive Bayes classifiers in determining both the right and left hand movement through electroencephalographic signals. To accomplish this, we considered as input feature the energy of the signal trials filtered by a band pass with sensorimotor rhythm frequencies; spectral power components from the Welch modified periodogram were also used. As a preprocessing form, the Common Spatial Pattern (CSP) filter was used to increase the discriminative activity between classes of movement. The database created from this study reached hit rates of up to 70% while the BCI Competition II reached hit rates up to 88%, which is consistent with the literature.

A Brain-Computer Interface (BCI) is a system to communicate with external world through the brain activity. The brain activity is measured by Electro-Encephalography (EEG) and then processed by a BCI system. EEG source reconstruction could be a way to improve the accuracy of EEG classification in EEGbased brain–computer interface (BCI). In this thesis BCI methods were applied on derived sources which by their EEG enhancement it became possible to obtain a more accurate EEG detection and brought a new application to BCI technology that are recognition of writing letters imagery from brain waves. The BCI system enables people to write and type letters by their brain activity (EEG). To this end, first part of the thesis is dedicated to EEG source reconstruction techniques to select the most optimal EEG channels for task classification purposes. Due to this reason the changes in EEG signal power from rest state to motor imagery task was used, to find the location of an active single equivalent dipole. Implementing an inverse problem solution on the power changes by Multiple Sparse Priors (MSP) method generated a scalp map where its fitting showed the localization of EEG electrodes. Having the optimized locations the secondary objective was to choose the most optimal EEG features and rhythm for an efficient classification. This became possible by feature ranking, 1- Nearest Neighbor leave-one-out. The feature vectors were computed by applying the combined methods of multitaper method, Pwelch. The features were classified by several methods of Normal densities based quadratic classifier (qdc), k-nearest neighbor classifier (knn), Mixture of Gaussians classification and Train neural network classifier using back-propagation. Results show that the selected features and classifiers are able to recognize the imagination of writing alphabet with the high accuracy. / BCI controls external devices and interacts with the environment by brain signals. Measured EEG signals over the motor cortex exhibit changes in power related to the movements or imaginations which are executed in motor tasks [1]. These changes declare increase or decrease of power in the alpha (8Hz-13Hz), and beta (13Hz-28Hz) frequency bands from resting state to motor imagery task that known as event related synchronization (in case of power increasing) and desynchronization (in case of power decreasing) [2]. The necessity to communicate with the external world for locked-in state (LIS) patients (a paralyzed patient who only communicates with eyes), made doctors and engineers motivated to develop a BCI technology for typing letters through brain commands. Many researches have been done around this area to ascertain the dream of typing for handicapped. In the brain some regions of the cerebral cortex (motor cortex) are involved in the planning, control, and execution of voluntary movements. Electroencephalography (EEG) signals are electrical potential generated by the nerve cells in the cerebral cortex. In order to execute motoric tasks, the EEG signals are appeared over the motor cortex [1]. The measured brain response to a stimulus is called eventrelated potential (ERP). P300-event related potential (ERP) is an evoked neuron response to an external auditory or visual stimulus that is detectable in scalp-recorded EEG (The P300 is evoked potential which occurs across the parieto-central on the skull 300 ms after applying the stimulus). Farwell and Donchin have proven in a P300-based BCI speller [3] that P300 response is a reliable signal for controlling a BCI system. They described the P300 speller, in which alphanumeric characters are represented in a matrix grid of six-by-six matrix. The user should focus on one of the 36 character cells while each row and column of the grid is intensified randomly and sequentially. The P300, observed in EEG signals, is created by the intersection of the target row and column which causes detection of the target stimuli with a probability of 1/6 (in case of high accuracy of flashing operation). Also when the target stimulus is rarely presented in the random sequence of stimuli causes a neural reaction to unpredictable but recognizable event and a P300 response is evoked [3]. Generally when the subject is involved with the task to recognize the targets, the P300 wave happens and the signal amplitude varies with the unlikelihood of the targets. Its dormancy changes with the difficulty of recognizing the target stimulus from the standard stimuli [3].The attended character of the matrix can be extracted by proper feature extraction and classification of P300. A plenty of procedures for feature extraction and classification have been applied to improve the performance of originally reported speller [3], such as stepwise linear discriminate analysis (SWLDA) [4, 5], wavelets [1], support vector machines [6, 7, 8] and matched filtering [9]. Till now, BCI-related P300 research has mostly considered on signals from standard P300 scalp locations. While in [10, 11, 12, 13, 14, 15, 16] it has been proven that the use of additional locations, especially posterior sites, may improve classification accuracy, but it has not been addressed to particular offline and online studies. Recently, auditory version improvement of the visual P300 speller allows locked in patients who have problem in the visual system to use the P300 speller system by relating two numbers to each letter which indicate the row and column of letter position [17]. Now a new technology is needed which can substitute a keyboard with no alphabet menu. The technology will be handy for blind people and useful for healthy persons who need to work hands free with their computer or mobile. The aim of this thesis is to improve EEG detection through source localization for a new BCI application to type with EEG signals without using alphabet menu. / +98-9359576229

Les Interfaces Cerveau-Machine (ICM) ont été appliquées avec succès aussi bien dans le domaine clinique que pour l'amélioration de la vie quotidienne de patients avec des handicaps. En tant que composante essentielle, le module de traitement du signal détermine nettement la performance d'un système ICM. Nous nous consacrons à améliorer les stratégies de traitement du signal du point de vue de l'apprentissage de la machine. Tout d'abord, nous avons développé un algorithme basé sur les SVM transductifs couplés aux noyaux multiples afin d'intégrer différentes vues des données (vue statistique ou vue géométrique) dans le processus d'apprentissage. Deuxièmement, nous avons proposé une version enligne de l'apprentissage multi-noyaux dans le cas supervisé. Les résultats expérimentaux montrent de meilleures performances par rapport aux approches classiques. De plus, l'algorithme proposé permet de sélectionner automatiquement les canaux de signaux EEG utiles grâce à l'apprentissage multi-noyaux.Dans la dernière partie, nous nous sommes attaqués à l'amélioration du module de traitement du signal au-delà des algorithmes d'apprentissage automatique eux-mêmes. En analysant les données ICM hors-ligne, nous avons d'abord confirmé qu'un modèle de classification simple peut également obtenir des performances satisfaisantes en effectuant une sélection de caractéristiques (et/ou de canaux). Nous avons ensuite conçu un système émotionnel ICM par en tenant compte de l'état émotionnel de l'utilisateur. Sur la base des données de l'EEG obtenus avec différents états émotionnels, c'est-à -dire, positives, négatives et neutres émotions, nous avons finalement prouvé que l'émotion affectait les performances ICM en utilisant des tests statistiques. Cette partie de la thèse propose des bases pour réaliser des ICM plus adaptées aux utilisateurs. / Brain-computer Interface (BCI) has achieved numerous successful applications in both clinicaldomain and daily life amelioration. As an essential component, signal processing determines markedly the performance of a BCI system. In this thesis, we dedicate to improve the signal processing strategy from perspective of machine learning strategy. Firstly, we proposed TSVM-MKL to explore the inputs from multiple views, namely, from statistical view and geometrical view; Secondly, we proposed an online MKL to reduce the computational burden involved in most MKL algorithm. The proposed algorithms achieve a better classifcation performance compared with the classical signal kernel machines, and realize an automatical channel selection due to the advantages of MKL algorithm. In the last part, we attempt to improve the signal processing beyond the machine learning algorithms themselves. We first confirmed that simple classifier model can also achieve satisfying performance by careful feature (and/or channel) selection in off-line BCI data analysis. We then implement another approach to improve the BCI signal processing by taking account for the user's emotional state during the signal acquisition procedure. Based on the reliable EEG data obtained from different emotional states, namely, positive, negative and neutral emotions, we perform strict evaluation using statistical tests to confirm that the emotion does affect BCI performance. This part of work provides important basis for realizing user-friendly BCIs.

Ce mémoire par article part de la question suivante: pouvons-nous utiliser des prothèses neurales afin d’activer artificiellement certain muscles dans le but d’accélérer la guérison et le réapprentissage du contrôle moteur après un AVC ou un traumatisme cervical ? Cette question touche plus de 15 millions de personnes chaque année à travers le monde, et est au coeur de la recherche de Numa Dancause et Marco Bonizzato, nos collaborateurs dans le département de Neuroscience de l’Université de Montréal. Il est maintenant possible d’implanter des électrodes à grande capacité dans le cortex dans le but d’acheminer des signaux électriques, mais encore difficile de prédire l’effet de stimulations sur le cerveau et le reste du corps. Cependant, des résultats préliminaires prometteurs sur des rats et singes démontrent qu’une récupération motrice non-négligeable est observée après stimulation de régions encore fonctionnelles du cortex moteur. Les difficultés rattachées à l’implémentation optimale de stimulation motocorticale consistent donc à trouver une de ces régions, ainsi qu’un protocole de stimulation efficace à la récupération. Bien que cette optimisation a été jusqu’à présent faite à la main, l’émergence d’implants capables de livrer des signaux sur plusieurs sites et avec plusieurs patrons spatio-temporels rendent l’exploration manuelle et exhaustive impossible. Une approche prometteuse afin d’automatiser et optimiser ce processus est d’utiliser un algorithme d’exploration bayésienne. Mon travail a été de déveloper et de raffiner ces techniques avec comme objectif de répondre aux deux questions scientifiques importantes suivantes: (1) comment évoquer des mouvements complexes en enchainant des microstimulations corticales ?, et (2) peuvent-elles avoir des effets plus significatifs que des stimulations simples sur la récupération motrice? Nous présentons dans l’article de ce mémoire notre approche hiérarchique utilisant des processus gaussiens pour exploiter les propriétés connues du cerveau afin d’accélérer la recherche, ainsi que nos premiers résultats répondant à la question 1. Nous laissons pour des travaux futur une réponse définitive à la deuxième question. / The idea for this thesis by article sprung from the following question: can we use neural prostheses to stimulate specific muscles in order to help recovery of motor control after stroke or cervical injury? This question is of crucial importance to 15 million people each year around the globe, and is at the heart of Numa Dancause and Marco Bonizzato’s research, our collaborators in the Neuroscience department at the University of Montreal. It is now possible to implant large capacity electrodes for electrical stimulation in cortex, but still difficult to predict their effect on the brain and the rest of the body. Nevertheless, preliminary but promising results on rats and monkeys have shown that a non-negligible motor recovery is obtained after stimulation of regions of motor cortex that are still functional. The difficulties related to optimal microcortical stimulation hence consist in finding both one of these regions, and a stimulation protocol with optimal recovery efficacy. This search has up to present day been performed by hand, but recent and upcoming large scale stimulation technologies permitting delivery of spatio-temporal signals are making such exhaustive searches impossible.A promising approach to automating and optimizing this discovery is the use of Bayesian optimization. My work has consisted in developing and refining such techniques with two scientific questions in mind: (1) how can we evoke complex movements by chaining cortical microstimulations?, and (2) can these outperform single channel stimulations in terms of recovery efficacy? We present in the main article of this thesis our hierarchical Bayesian optimization approach which uses gaussian processes to exploit known properties of the brain to speed up the search, as well as first results answering question 1. We leave to future work a definitive answer to the second question.