Systems, Man & Cybernetics - October 2017 - 18

FP1
FP2
F7
F3
FZ

Convenient Signal
Recording

F3
FZ

Signal Quality
Enhancement

Signal
Preprocessing

Signal
Acquisition

User

FP1
FP2

BCI
Signal
Translation
Applications

Actionable
Decision Fusion

Feature
Extraction
Brain Network
Measurement

Figure 1. Current neurotechnology and computational intelligence methods applied to enhance BCI performance.

optimal solution for a multimodal approach because its
signals do not interfere with electric or magnetic fields
[20]. There are several types of EEG signals used to
design and operate BCIs, such as P300 ERPs and steadystate visually evoked potentials (SSVEPs). Among them,
P300 and SSVEP signals have become extremely popular
due to the high information transfer rate (ITR) they produce and their minimal user training requirement [21],
[22]. Each collected signal possesses its own properties
and potential uncertainties to describe the underlying cognitive states. A comprehensive analysis of multiple sources is needed to reduce individual uncertainty and improve
the system performance reliability. Therefore, developing
an effective approach to integrate multimodal information
is an important and urgent issue.
In this article, current neurotechnology and computational intelligence methods are introduced as possible solutions to address the aforementioned technical
issues. In contrast to conventional wet electrodes, dry
electrodes exhibit the electronic characteristics of electrically conductive materials. They obtain high-quality
signals without skin abrasion or preparation. In the
novel capture EEG devices presented in this article, dry
electrodes act as substitutes for traditional wet electrodes; these dry electrodes can acquire real-time EEG
signals for operational workplaces without requiring
conductive gel/paste or scalp preparation in BCI applications. Additionally, an online artifact removal
18

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE Oc tob e r 2017

technique based on canonical correlation analysis
(CCA) [13] as a BSS used to remove artifacts is presented in this article. The feasibility of rapid P3 detection using a radial basis function network (RBFN) [23]
under a small number of EEG trials is also demonstrated. Granger causality (GC) analysis is introduced to
extract detailed changes in the brain network during
sustained-attention driving, and Dempster-Shafer (D-S)
theory [24] and [25] is used to aggregate pieces of evidence from multiple information sources and exploit
redundancy and complementariness between sources
in global information. When applied to integrate different physiological signals, this fusion technique can
improve the quality of final decisions and facilitate the
optimal estimation of objects.
EEG-Based Neuroimaging Technology for BCIs
Conventional wet electrodes are commonly used to measure EEG signals. These electrodes provide excellent
EEG signals with the proper skin preparation and conductive gel application; however, the skin must be prepared prior to applying the wet electrodes, which is
typically problematic for users. To overcome these drawbacks, we developed several types of novel dry-contact
EEG sensors that can efficiently reduce preparation time
without conductive gel. Figure 2 shows several types of
dry sensors, including spring-loaded [26], foam-based
[27], and silicon-based sensors [28]. The dr y foam

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - October 2017