Tracey A. Camilleri

Automatic detection of spindles and K-complexes in sleep EEG using switching multiple models

Abstract This work investigates the use of switching linear Gaussian state space models for the segmentation and automatic labelling of Stage 2 sleep EEG data characterised by spindles and K-complexes. The advantage of this approach is that it offers a unified framework of detecting multiple transient events within background EEG data. Specifically for the identification of background EEG, spindles and K-complexes, a true positive rate (false positive rate) of 76.04% (33.47%), 83.49% (47.26%) and 52.02% (7.73%) respectively was obtained on a sample by sample basis. A novel semi-supervised model allocation approach is also proposed, allowing new unknown modes to be learnt in real time.

Interfacing with a speller using EOG glasses

Bio-signal based human computer interface (HCI) systems are a good alternative to standard touch based interfaces, offering subjects with motor impairments an alternative means of communication. This work investigates the use of electrooculography (EOG) to interface with a speller application. The use of a wireless EOG glasses currently on the market, known as JINS MEME, comprising only three dry electrodes, is compared to the standard two-pair EOG electrode configuration using wet electrodes. A blink accuracy of 97.63% and a saccade accuracy of 73.38% was obtained using a novel thresholding algorithm on the EOG data collected through the MEME glasses and the results were shown to be comparable to those obtained using wet surface electrodes. A real-time menu driven keyboard is also proposed and tested using the different eye movement recording techniques. In this case an average writing speed of 7.11 letters per minute, with a classification accuracy of 90.59% was obtained using signals recorded from the MEME glasses, showing that this new technology offers an ergonomic system that can easily be used in eye-based assistive applications.

Comparison of plain and checkerboard stimuli for brain computer interfaces based on steady state visual evoked potentials

A steady-state visual evoked potential (SSVEP) is a neural response observed in the visual cortex evoked by repetitive visual stimulation. In an SSVEP-based brain-computer interface (BCI) application various visual stimuli that induce SSVEPs at different frequencies are associated with distinct commands; a user activates a particular command by focusing on the targeted stimulus. The pattern of these visual stimuli is one of the properties that affects the accuracy of SSVEP detection. This work thus compares the two most common types of stimuli, which are the plain and checkerboard stimuli, for BCI systems. Results showed a statistically significant 9.26% average increase in SSVEP classification when using a plain stimulus over a checkerboard stimulus.

EEG-based biometry using steady state visual evoked potentials

The use of brain signals for person recognition has in recent years attracted considerable interest because of the increased security and privacy these can offer when compared to conventional biometric measures. The main challenge lies in extracting features from the EEG signals that are sufficiently distinct across individuals while also being sufficiently consistent across multiple recording sessions. A range of EEG phenomena including eyes open and eyes closed activity, visual evoked potentials (VEPs) through image presentation, and other mental tasks have been studied for their use in biometry.

An Analysis on the Effect of Phase on the Performance of SSVEP-based BCIs

In recent years, brain-computer interface (BCI) systems have emerged as a technology that can support a direct communication pathway between the human brain and external devices. Among the various neurophysiological phenomena that can be used to drive BCI systems, steady-state visual evoked potentials (SSVEPs) have gained increasing popularity because of the high information transfer rate (ITR) and high accuracy that these can provide. In this paper, an investigation on how the inclusion of phase information in the form of a novel proposed feature, the phase-weighted SNR (PWS) feature, can improve the performance of such setups is presented. Improvements in classification accuracies of up to 17% were obtained with the inclusion of phase information when compared to systems that rely solely on amplitude information.

Semi-supervised segmentation of EEG data in BCI systems

This work investigates the use of a semi-supervised, autoregressive switching multiple model (AR-SMM) framework for the segmentation of EEG data applied to brain computer interface (BCI) systems. This gives the possibility of identifying and learning novel modes within the data, giving insight on the changing dynamics of the EEG data and possibly also offering a solution for shorter training periods in BCIs. Furthermore it is shown that the semi-supervised model allocation process is robust to different starting positions and gives consistent results.

Segmentation and Labelling of EEG for Brain Computer Interfaces

Segmentation and labelling of time series is a common requirement for several applications. A brain computer interface (BCI) is achieved by classification of time intervals of the electroencephalographic (EEG) signal and thus requires EEG signal segmentation and labelling. This work investigates the use of an autoregressive model, extended to a switching multiple modelling framework, to automatically segment and label EEG data into distinct modes of operation that may switch abruptly and arbitrarily in time. The applicability of this approach to BCI systems is illustrated on an eye closure dependent BCI and on a motor imagery based BCI. Results show that the proposed autoregressive switching multiple model approach offers a unified framework of detecting multiple modes, even in the presence of limited training data.

EOG-based eye movement detection and gaze estimation for an asynchronous virtual keyboard

Abstract This work aims to develop a novel electrooculography (EOG)-based virtual keyboard with a standard QWERTY layout which, unlike similar state-of-the-art systems, allows users to reach any icon from any location directly and asynchronously. The saccadic EOG potential displacement is mapped to angular gaze displacement using a novel two-channel input linear regression model, which considers features extracted from both the horizontal and vertical EOG signal components jointly. Using this technique, a gaze displacement estimation error of 1.32 ± 0.26° and 1.67 ± 0.26° in the horizontal and vertical directions respectively was achieved, a performance which was also found to be generally statistically significantly better than the performance obtained using one model for each EOG component to model the relationship in the horizontal and vertical directions separately, as typically used in the literature. Furthermore, this work also proposes a threshold-based method to detect eye movements from EOG signals in real-time, which are then classified as saccades or blinks using a novel cascade of a parametric and a signal-morphological classifier based on the EOG peak and gradient features. This resulted in an average saccade and blink labelling accuracy of 99.92% and 100.00% respectively, demonstrating that these two eye movements could be reliably detected and discriminated in real-time using the proposed algorithms. When these techniques were used to interface with the proposed asynchronous EOG-based virtual keyboard, an average writing speed across subjects of 11.89 ± 4.42 characters per minute was achieved, a performance which has been shown to improve substantially with user experience.

A systematic review investigating the relationship of electroencephalography and magnetoencephalography measurements with sensorimotor upper limb impairments after stroke

BACKGROUND Predicting sensorimotor upper limb outcome receives continued attention in stroke. Neurophysiological measures by electroencephalography (EEG) and magnetoencephalography (MEG) could increase the accuracy of predicting sensorimotor upper limb recovery. NEW METHOD The aim of this systematic review was to summarize the current evidence for EEG/MEG-based measures to index neural activity after stroke and the relationship between abnormal neural activity and sensorimotor upper limb impairment. Relevant papers from databases EMBASE, CINHAL, MEDLINE and pubMED were identified. Methodological quality of selected studies was assessed with the Modified Downs and Black form. Data collected was reported descriptively. RESULTS Seventeen papers were included; 13 used EEG and 4 used MEG applications. Findings showed that: (a) the presence of somatosensory evoked potentials in the acute stage are related to better outcome of upper limb motor impairment from 10 weeks to 6 months post-stroke; (b) an interhemispheric imbalance of cortical oscillatory signals associated with upper limb impairment; and (c) predictive models including beta oscillatory cortical signal factors with corticospinal integrity and clinical measures could enhance upper limb motor prognosis. COMPARING WITH EXISTING METHOD The combination of neurological biomarkers with clinical measures results in higher statistical power than using neurological biomarkers alone when predicting motor recovery in stroke. CONCLUSIONS Alterations in neural activity by means of EEG and MEG are demonstrated from the early post-stroke stage onwards, and related to sensorimotor upper limb impairment. Future work exploring cortical oscillatory signals in the acute stage could provide further insight about prediction of upper limb sensorimotor recovery.

To train or not to train? A survey on training of feature extraction methods for SSVEP-based BCIs

OBJECTIVE Despite the vast research aimed at improving the performance of steady-state visually evoked potential (SSVEP)-based brain-computer interfaces (BCIs), several limitations exist that restrict the use of such applications for long-term users in the real-world. One of the main challenges has been to reduce training time while maintaining good BCI performance. In view of this challenge, this survey identifies and compares the different training requirements of feature extraction methods for SSVEP-based BCIs. APPROACH This paper reviews the various state-of-the-art SSVEP feature extraction methods that have been developed and are most widely used in the literature. MAIN RESULTS The main contributions compared to existing reviews are the following: (i) a detailed summary, including a brief mathematical description of each feature extraction algorithm, providing a guide to the basic concepts of the state-of-the-art techniques for SSVEP-based BCIs found in literature; (ii) a categorisation of the training requirements of SSVEP-based methods into three categories, defined as training-free methods, subject-specific and subject-independent training methods; (iii) a comparative review of the training requirements of SSVEP feature extraction methods, providing a reference for future work on SSVEP-based BCIs. SIGNIFICANCE This review highlights the strengths and weaknesses of the three categories of SSVEP training methods. Training-free systems are more practical but their performance is limited due to inter-subject variability resulting from the complex EEG activity. Feature extraction methods that incorporate some training data address this issue and in fact have outperformed training-free methods: subject-specific BCIs are tuned to the individual yielding the best performance at the cost of long, tiring training sessions making these methods unsuitable for everyday use; subject-independent BCIs that make use of training data from various subjects offer a good trade-off between training effort and performance, making these BCIs better suited for practical use.