Brain computer interface (BCI) systems measure brain signal and translate it into control commands in an attempt to mimic specific human thinking activities. In recent years, many researchers have shown their interests in BCI systems, which has resulted in many experiments and applications. However, most methods are just based on a specific selected dataset or a typical feature. As a result, there are questions about whether some methods generalise well on other datasets. Therefore, the major motivation of this thesis is to compare various features and classifiers described in the literature. Pattern recognition is considered as the core part of a BCI system in our research. In this thesis, a number of different features and classifiers are compared in terms of classification accuracy and computation time. The studied features are: time series waveform, autoregressive (AR) components, spectral components; these are used with different classifiers: such as template matching, nearest neighbour, linear discriminant analysis (LDA), Bayesian statistical and fuzzy logic decision classifiers. In order to assess and compare these different features and classifiers, an extensive investigation was carried out on a public dataset (imagined left or right hand movement) from an international BCI competition and the results are reported in this thesis. The classification was done in a continuous fashion, to match a real time application. In this process, the average and best accuracy, as well as the computation time, were analysed and compared. The results showed that most classifiers achieved very high accuracies and short computation times for most features. A BCI experiment based on imagined left or right hand movement was carried out at the University of Adelaide and some investigations on the data from this experiment are discussed. The result shows that the selected classifiers can work well with this new dataset without much additional preprocessing or modifications. Finally, this thesis culminates with some conclusions based on our research, and discusses some further potential work. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1415396 / Thesis (M.Eng.Sc.) - University of Adelaide, School of Electrical and Electronic Engineering, 2010

P300 spellers can provide a means of communication for individuals with severe neuromuscular limitations. However, its use as an effective communication tool is reliant on high P300 classification accuracies (>70‰) to account for error revisions. Error-related potentials (ErrP), which are changes in EEG potentials when a person is aware of or perceives erroneous behavior or feedback, have been proposed as inputs to drive corrective mechanisms that veto erroneous actions by BCI systems. The goal of this study is to demonstrate that training an additional ErrP classifier for a P300 speller is not necessary, as we hypothesize that error information is encoded in the P300 classifier responses used for character selection. We perform offline simulations of P300 spelling to compare ErrP and non-ErrP based corrective algorithms. A simple dictionary correction based on string matching and word frequency significantly improved accuracy (35-185%), in contrast to an ErrP-based method that flagged, deleted and replaced erroneous characters (-47-0‰). Providing additional information about the likelihood of characters to a dictionary-based correction further improves accuracy. Our Bayesian dictionary-based correction algorithm that utilizes P300 classifier confidences performed comparably (44-416%) to an oracle ErrP dictionary-based method that assumed perfect ErrP classification (43-433%).

An EEG-based brain-computer system can be used to control external devices such as computers, wheelchairs or Virtual Environments. One of the most important applications is a spelling device to aid severely disabled individuals with communication, for example people disabled by amyotrophic lateral sclerosis (ALS). P300-based BCI systems are optimal for spelling characters with high speed and accuracy, as compared to other BCI paradigms such as motor imagery. In this study, 100 subjects tested a P300-based BCI system to spell a 5-character word with only 5 min of training. EEG data were acquired while the subject looked at a 36-character matrix to spell the word WATER. Two different versions of the P300 speller were used: (i) the row/column speller (RC) that flashes an entire column or row of characters and (ii) a single character speller (SC) that flashes each character individually. The subjects were free to decide which version to test. Nineteen subjects opted to test both versions. The BCI system classifier was trained on the data collected for the word WATER. During the real-time phase of the experiment, the subject spelled the word LUCAS, and was provided with the classifier selection accuracy after each of the five letters. Additionally, subjects filled out a questionnaire about age, sex, education, sleep duration, working duration, cigarette consumption, coffee consumption, and level of disturbance that the flashing characters produced. 72.8% (N = 81) of the subjects were able to spell with 100% accuracy in the RC paradigm and 55.3% (N = 38) of the subjects spelled with 100% accuracy in the SC paradigm. Less than 3% of the subjects did not spell any character correctly. People who slept less than 8 h performed significantly better than other subjects. Sex, education, working duration, and cigarette and coffee consumption were not statistically related to differences in accuracy. The disturbance of the flashing characters was rated with a median score of 1 on a scale from 1 to 5 (1, not disturbing; 5, highly disturbing). This study shows that high spelling accuracy can be achieved with the P300 BCI system using approximately 5 min of training data for a large number of non-disabled subjects, and that the RC paradigm is superior to the SC paradigm. 89% of the 81 RC subjects were able to spell with accuracy 80-100%. A similar study using a motor imagery BCI with 99 subjects showed that only 19% of the subjects were able to achieve accuracy of 80-100%. These large differences in accuracy suggest that with limited amounts of training data the P300-based BCI is superior to the motor imagery BCI. Overall, these results are very encouraging and a similar study should be conducted with subjects who have ALS to determine if their accuracy levels are similar.

In recent years, brain computer interfaces (BCIs) have gained popularity in non-medical domains such as the gaming, entertainment, personal health, and marketing industries. A growing number of companies offer various inexpensive consumer grade BCIs and some of these companies have recently introduced the concept of BCI "App stores" in order to facilitate the expansion of BCI applications and provide software development kits (SDKs) for other developers to create new applications for their devices. The BCI applications access to users' unique brainwave signals, which consequently allows them to make inferences about users' thoughts and mental processes. Since there are no specific standards that govern the development of BCI applications, its users are at the risk of privacy breaches. In this work, we perform first comprehensive analysis of BCI App stores including software development kits (SDKs), application programming interfaces (APIs), and BCI applications w.r.t privacy issues. The goal is to understand the way brainwave signals are handled by BCI applications and what threats to the privacy of users exist. Our findings show that most applications have unrestricted access to users' brainwave signals and can easily extract private information about their users without them even noticing. We discuss potential privacy threats posed by current practices used in BCI App stores and then describe some countermeasures that could be used to mitigate the privacy threats. Also, develop a prototype which gives the BCI app users a choice to restrict their brain signal dynamically.

Brain-computer interfaces (BCIs) might restore communication to people severely disabled by amyotrophic lateral sclerosis (ALS) or other disorders. We sought to: 1) define a protocol for determining whether a person with ALS can use a visual P300-based BCI; 2) determine what proportion of this population can use the BCI; and 3) identify factors affecting BCI performance. Twenty-five individuals with ALS completed an evaluation protocol using a standard 6 × 6 matrix and parameters selected by stepwise linear discrimination. With an 8-channel EEG montage, the subjects fell into two groups in BCI accuracy (chance accuracy 3%). Seventeen averaged 92 (± 3)% (range 71-100%), which is adequate for communication (G70 group). Eight averaged 12 (± 6)% (range 0-36%), inadequate for communication (L40 subject group). Performance did not correlate with disability: 11/17 (65%) of G70 subjects were severely disabled (i.e. ALSFRS-R < 5). All L40 subjects had visual impairments (e.g. nystagmus, diplopia, ptosis). P300 was larger and more anterior in G70 subjects. A 16-channel montage did not significantly improve accuracy. In conclusion, most people severely disabled by ALS could use a visual P300-based BCI for communication. In those who could not, visual impairment was the principal obstacle. For these individuals, auditory P300-based BCIs might be effective.

Approximately 20 million people in the United States suffer from irreversible nerve damage and would benefit from a neuroprosthetic device modulated by a Brain-Computer Interface (BCI). These devices restore independence by replacing peripheral nervous system functions such as peripheral control. Although there are currently devices under investigation, contemporary methods fail to offer adaptability and proper signal recognition for output devices. Human anatomical differences prevent the use of a fixed model system from providing consistent classification performance among various subjects. Furthermore, notoriously noisy signals such as Electroencephalography (EEG) require complex measures for signal detection. Therefore, there remains a tremendous need to explore and improve new algorithms. This report investigates a signal-processing model that is better suited for BCI applications because it incorporates machine learning and fuzzy logic. Whereas traditional machine learning techniques utilize precise functions to map the input into the feature space, fuzzy-neuro system apply imprecise membership functions to account for uncertainty and can be updated via supervised learning. Thus, this method is better equipped to tolerate uncertainty and improve performance over time. Moreover, a variation of this algorithm used in this study has a higher convergence speed. The proposed two-stage signal-processing model consists of feature extraction and feature translation, with an emphasis on the latter. The feature extraction phase includes Blind Source Separation (BSS) and the Discrete Wavelet Transform (DWT), and the feature translation stage includes the Takagi-Sugeno-Kang Fuzzy-Neural Network (TSKFNN). Performance of the proposed model corresponds to an average classification accuracy of 79.4 % for 40 subjects, which is higher than the standard literature values, 75%, making this a superior model.

The objective of this study is to test whether human intentions to sustain or cease movements in right and left hands can be decoded reliably from spatially filtered single trial magneto-encephalographic (MEG) signals. This study was performed using motor execution and motor imagery movements to achieve a potential high performance Brain-Computer interface (BCI). Seven healthy volunteers, naïve to BCI technology, participated in this study. Signals were recorded from 275-channel MEG and synthetic aperture magnetometry (SAM) was employed as the spatial filter. The four-class classification for natural movement intentions was performed offline; Genetic Algorithm based Mahalanobis Linear Distance (GA-MLD) and direct-decision tree classifier (DTC) techniques were adopted for the classification through 10-fold cross-validation. Through SAM imaging, strong and distinct event related desynchronisation (ERD) associated with sustaining, and event related synchronisation (ERS) patterns associated with ceasing of hand movements were observed in the beta band (15 - 30 Hz). The right and left hand ERD/ERS patterns were observed on the contralateral hemispheres for motor execution and motor imagery sessions. Virtual channels were selected from these cortical areas of high activity to correspond with the motor tasks as per the paradigm of the study. Through a statistical comparison between SAM-filtered virtual channels from single trial MEG signals and basic MEG sensors, it was found that SAM-filtered virtual channels significantly increased the classification accuracy for motor execution (GA-MLD: 96.51 ± 2.43 %) as well as motor imagery sessions (GA-MLD: 89.69 ± 3.34%). Thus, multiple movement intentions can be reliably detected from SAM-based spatially-filtered single trial MEG signals. MEG signals associated with natural motor behavior may be utilized for a reliable high-performance brain-computer interface (BCI) and may reduce long-term training compared with conventional BCI methods using rhythm control. This may prove tremendously helpful for patients suffering from various movement disorders to improve their quality of life.

In this study, a P300 based Brain-Computer Interface (BCI) system design is realized by the implementation of the Spelling Paradigm. The main challenge in these systems is to improve the speed of the prediction mechanisms by the application of different signal processing and pattern classification techniques in BCI problems. The thesis study includes the design and implementation of a 10 channel Electroencephalographic (EEG) data acquisition system to be practically used in BCI applications. The electrical measurements are realized with active electrodes for continuous EEG recording. The data is transferred via USB so that the device can be operated by any computer. v Wiener filtering is applied to P300 Speller as a signal enhancement tool for the first time in the literature. With this method, the optimum temporal frequency bands for user specific P300 responses are determined. The classification of the responses is performed by using Support Vector Machines (SVM&rsquo / s) and Bayesian decision. These methods are independently applied to the row-column intensification groups of P300 speller to observe the differences in human perception to these two visual stimulation types. It is observed from the investigated datasets that the prediction accuracies in these two groups are different for each subject even for optimum classification parameters. Furthermore, in these datasets, the classification accuracy was improved when the signals are preprocessed with Wiener filtering. With this method, the test characters are predicted with 100% accuracy in 4 trial repetitions in P300 Speller dataset of BCI Competition II. Besides, only 8 trials are needed to predict the target character with the designed BCI system.

This thesis study focuses on the realization of an online cue based Motor Imagery (MI) Brain Computer Interface (BCI). For this purpose, some signal processing and classification methods are investigated. Specifically, several time-spatial-frequency methods, namely the Short Time Fourier Transform (STFT), Common Spatial Frequency Patterns (CSFP) and the Morlet Transform (MT) are implemented on a 2-class MI BCI system. Distinction Sensitive Learning Vector Quantization (DSLVQ) method is used as a feature selection method. The performance of these methodologies is evaluated with the linear and nonlinear Support Vector Machines (SVM), Multilayer Perceptron (MLP) and Naive Bayesian (NB) classifiers. The methodologies are tested on BCI Competition IV dataset IIb and an average kappa value of 0.45 is obtained on the dataset. According to the classification results, the algorithms presented here obtain the 4th level in the competition as compared to the other algorithms in the competition. Offline experiments are performed in METU Brain Research Laboratories and Hacettepe Biophysics Department on two subjects with the original cue-based MI BCI paradigm. Average prediction accuracy of the methods on a 2-class BCI is evaluated to be 76.26% in these datasets. Furthermore, two online BCI applications are developed: the ping-pong game and the electrical wheelchair control. For these applications, average classification accuracy is found to be 70%. During the offline experiments, the performance of the developed system is observed to be highly dependent on the subject training and experience. According to the results, the EEG channels P3 and P4, which are considered to be irrelevant with the motor imagination, provided the best classification performance on the offline experiments. Regarding the observations on the experiments, this process is related to the stimulation mechanism in the cue based applications and consequent visual evoking effects on the subjects.

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.