Zhenghui Gu

Robust Adaptive Beamformers Based on Worst-Case Optimization and Constraints on Magnitude Response

In this paper, novel robust adaptive beamformers are proposed with constraints on array magnitude response. With the transformation from the array output power and the magnitude response to linear functions of the autocorrelation sequence of the array weight, the optimization of an adaptive beamformer, which is often described as a quadratic optimization problem in conventional beamforming methods, is then reformulated as a linear programming (LP) problem. Unlike conventional robust beamformers, the proposed method is able to flexibly control the robust response region with specified beamwidth and response ripple. In practice, an array has many imperfections besides steering direction error. In order to make the adaptive beamformer robust against all kinds of imperfections, worst-case optimization is exploited to reconstruct the robust beamformer. By minimizing array output power with the existence of the worst-case array imperfections, the robust beamforming can be expressed as a second-order cone programming (SOCP) problem. The resultant beamformer possesses superior robustness against arbitrary array imperfections. With the proposed methods, a large robust response region and a high signal-to-interference-plus-noise ratio (SINR) enhancement can be achieved readily. Simple implementation, flexible performance control, as well as significant SINR enhancement, support the practicability of the proposed methods.

Voxel Selection in fMRI Data Analysis Based on Sparse Representation

Multivariate pattern analysis approaches toward detection of brain regions from fMRI data have been gaining attention recently. In this study, we introduce an iterative sparse-representation-based algorithm for detection of voxels in functional MRI (fMRI) data with task relevant information. In each iteration of the algorithm, a linear programming problem is solved and a sparse weight vector is subsequently obtained. The final weight vector is the mean of those obtained in all iterations. The characteristics of our algorithm are as follows: 1) the weight vector (output) is sparse; 2) the magnitude of each entry of the weight vector represents the significance of its corresponding variable or feature in a classification or regression problem; and 3) due to the convergence of this algorithm, a stable weight vector is obtained. To demonstrate the validity of our algorithm and illustrate its application, we apply the algorithm to the Pittsburgh Brain Activity Interpretation Competition 2007 functional fMRI dataset for selecting the voxels, which are the most relevant to the tasks of the subjects. Based on this dataset, the aforementioned characteristics of our algorithm are analyzed, and a comparison between our method with the univariate general-linear-model-based statistical parametric mapping is performed. Using our method, a combination of voxels are selected based on the principle of effective/sparse representation of a task. Data analysis results in this paper show that this combination of voxels is suitable for decoding tasks and demonstrate the effectiveness of our method.

Target Selection With Hybrid Feature for BCI-Based 2-D Cursor Control

To control a cursor on a monitor screen, a user generally needs to perform two tasks sequentially. The first task is to move the cursor to a target on the monitor screen (termed a 2-D cursor movement), and the second task is either to select a target of interest by clicking on it or to reject a target that is not of interest by not clicking on it. In a previous study, we implemented the former function in an EEG-based brain-computer interface system using motor imagery and the P300 potential to control the horizontal and vertical cursor movements, respectively. In this study, the target selection or rejection functionality is implemented using a hybrid feature from motor imagery and the P300 potential. Specifically, to select the target of interest, the user must focus his or her attention on a flashing button to evoke the P300 potential, while simultaneously maintaining an idle state of motor imagery. Otherwise, the user performs left-/right-hand motor imagery without paying attention to any buttons to reject the target. Our data analysis and online experimental results validate the effectiveness of our approach. The proposed hybrid feature is shown to be more effective than the use of either the motor imagery feature or the P300 feature alone. Eleven subjects attended our online experiment, in which a trial involved sequential 2-D cursor movement and target selection. The average duration of each trial and average accuracy of target selection were 18.19 s and 93.99% , respectively, and each target selection or rejection event was performed within 2 s.

Surfing the internet with a BCI mouse

In this paper, we present a new web browser based on a two-dimensional (2D) brain-computer interface (BCI) mouse, where our major concern is the selection of an intended target in a multi-target web page. A real-world web page may contain tens or even hundreds of targets, including hyperlinks, input elements, buttons, etc. In this case, a target filter designed in our system can be used to exclude most of those targets of no interest. Specifically, the user filters the targets of no interest out by inputting keywords with a P300-based speller, while keeps those containing the keywords. Such filtering largely facilitates the target selection task based on our BCI mouse. When there are only several targets in a web page (either an original sparse page or a target-filtered page), the user moves the mouse toward the target of interest using his/her electroencephalographic signal. The horizontal movement and vertical movement are controlled by motor imagery and P300 potential, respectively. If the mouse encounters a target of no interest, the user rejects it and continues to move the mouse. Otherwise the user selects the target and activates it. With the collaboration of the target filtering and a series of mouse movements and target selections/rejections, the user can select an intended target in a web page. Based on our browser system, common navigation functions, including history rolling forward and backward, hyperlink selection, page scrolling, text input, etc, are available. The system has been tested on seven subjects. Experimental results not only validated the efficacy of the proposed method, but also showed that free internet surfing with a BCI mouse is feasible.

Control of a Wheelchair in an Indoor Environment Based on a Brain–Computer Interface and Automated Navigation

The concept of controlling a wheelchair using brain signals is promising. However, the continuous control of a wheelchair based on unstable and noisy electroencephalogram signals is unreliable and generates a significant mental burden for the user. A feasible solution is to integrate a brain-computer interface (BCI) with automated navigation techniques. This paper presents a brain-controlled intelligent wheelchair with the capability of automatic navigation. Using an autonomous navigation system, candidate destinations and waypoints are automatically generated based on the existing environment. The user selects a destination using a motor imagery (MI)-based or P300-based BCI. According to the determined destination, the navigation system plans a short and safe path and navigates the wheelchair to the destination. During the movement of the wheelchair, the user can issue a stop command with the BCI. Using our system, the mental burden of the user can be substantially alleviated. Furthermore, our system can adapt to changes in the environment. Two experiments based on MI and P300 were conducted to demonstrate the effectiveness of our system.

Discrimination Between Control and Idle States in Asynchronous SSVEP-Based Brain Switches: A Pseudo-Key-Based Approach

A steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) can operate as an asynchronous brain switch. When SSVEP is detected with the “on/off” button flickering at a fixed frequency, the subject is identified as in the control state. Otherwise, he is in the idle state. Generally, the detection of the idle/control state is based on a predefined threshold, which is related to power. However, due to the variability of the electroencephalogram (EEG) signal, it is difficult to find an optimal threshold to achieve a high true-positive rate (TPR) in the control state while maintaining a low false-positive rate (FPR) in the idle state. In this paper, a novel pseudo-key-based approach is presented for better discriminating the control and idle states. A dedicated “on/off” button (target key) and several additional buttons (pseudo-keys) are displayed on the graphical user interface (GUI), and all of these buttons flash at different frequencies. The control state is identified from the EEG signal under two conditions. The first is a common thresholding condition, where the power ratio of the target key frequency component to a certain neighboring frequency band is above a predefined threshold. The second is a comparison condition, where the power of the target key frequency component is higher than any of the pseudo-keys. The effectiveness of the proposed approach is validated by several experiments. Further analysis shows that introducing the pseudo-keys can significantly reduce the probability that the SSVEP will be detected in response to the flickering target key in the idle state without substantially affecting the detection in the control state, providing strong evidence in support of our approach.

An asynchronous wheelchair control by hybrid EEG–EOG brain–computer interface

Wheelchair control requires multiple degrees of freedom and fast intention detection, which makes electroencephalography (EEG)-based wheelchair control a big challenge. In our previous study, we have achieved direction (turning left and right) and speed (acceleration and deceleration) control of a wheelchair using a hybrid brain–computer interface (BCI) combining motor imagery and P300 potentials. In this paper, we proposed hybrid EEG-EOG BCI, which combines motor imagery, P300 potentials, and eye blinking to implement forward, backward, and stop control of a wheelchair. By performing relevant activities, users (e.g., those with amyotrophic lateral sclerosis and locked-in syndrome) can navigate the wheelchair with seven steering behaviors. Experimental results on four healthy subjects not only demonstrate the efficiency and robustness of our brain-controlled wheelchair system but also indicate that all the four subjects could control the wheelchair spontaneously and efficiently without any other assistance (e.g., an automatic navigation system).

Decoding hand movement velocity from electroencephalogram signals during a drawing task

BackgroundDecoding neural activities associated with limb movements is the key of motor prosthesis control. So far, most of these studies have been based on invasive approaches. Nevertheless, a few researchers have decoded kinematic parameters of single hand in non-invasive ways such as magnetoencephalogram (MEG) and electroencephalogram (EEG). Regarding these EEG studies, center-out reaching tasks have been employed. Yet whether hand velocity can be decoded using EEG recorded during a self-routed drawing task is unclear.MethodsHere we collected whole-scalp EEG data of five subjects during a sequential 4-directional drawing task, and employed spatial filtering algorithms to extract the amplitude and power features of EEG in multiple frequency bands. From these features, we reconstructed hand movement velocity by Kalman filtering and a smoothing algorithm.ResultsThe average Pearson correlation coefficients between the measured and the decoded velocities are 0.37 for the horizontal dimension and 0.24 for the vertical dimension. The channels on motor, posterior parietal and occipital areas are most involved for the decoding of hand velocity. By comparing the decoding performance of the features from different frequency bands, we found that not only slow potentials in 0.1-4 Hz band but also oscillatory rhythms in 24-28 Hz band may carry the information of hand velocity.ConclusionsThese results provide another support to neural control of motor prosthesis based on EEG signals and proper decoding methods.

Sparse Representation for Brain Signal Processing: A tutorial on methods and applications

In many cases, observed brain signals can be assumed as the linear mixtures of unknown brain sources/components. It is the task of blind source separation (BSS) to find the sources. However, the number of brain sources is generally larger than the number of mixtures, which leads to an underdetermined model with infinite solutions. Under the reasonable assumption that brain sources are sparse within a domain, e.g., in the spatial, time, or time-frequency domain, we may obtain the sources through sparse representation. As explained in this article, several other typical problems, e.g., feature selection in brain signal processing, can also be formulated as the underdetermined linear model and solved by sparse representation. This article first reviews the probabilistic results of the equivalence between two important sparse solutions - the 0-norm and 1-norm solutions. In sparse representation-based brain component analysis including blind separation of brain sources and electroencephalogram (EEG) inverse imaging, the equivalence is related to the recoverability of the sources. This article also focuses on the applications of sparse representation in brain signal processing, including components extraction, BSS and EEG inverse imaging, feature selection, and classification. Based on functional magnetic resonance imaging (fMRI) and EEG data, the corresponding methods and experimental results are reviewed.

A Robust Adaptive Beamformer Based on Worst-Case Semi-Definite Programming

In this correspondence, a novel robust adaptive beamformer is proposed based on the worst-case semi-definite programming (SDP). A recent paper has reported that a beamformer robust against large steering direction error can be constructed by using linear constraints on magnitude response in SDP formulation. In practice, however, array system also suffers from many other array imperfections other than steering direction error. In order to make the adaptive beamformer robust against all kinds of array imperfections, the worst-case optimization technique is proposed to reformulate the beamformer by minimizing the array output power with respect to the worst-case array imperfections. The resultant beamformer has the mathematical form of a regularized SDP problem and possesses superior robustness against arbitrary array imperfections. Although the formulation of robust beamformer uses weighting matrix, with the help of spectral factorization approach, the weighting vector can be obtained so that the beamformer can be used for both signal power and waveform estimation. Simple implementation, flexible performance control, as well as significant signal-to-interference-plus-noise ratio (SINR) enhancement, support the practicability of the proposed method.