Current Source Density Estimation Enhances the Performance of Motor-Imagery-Related Brain–Computer Interface

Abstract

The objective is to evaluate the impact of EEG referencing schemes and spherical surface Laplacian (SSL) methods on the classification performance of motor-imagery (MI)-related brain-computer interface systems. Two EEG referencing schemes: common referencing and common average referencing and three surface Laplacian methods: current source density (CSD), finite difference method, and SSL using realistic head model were implemented separately for pre-processing of the EEG signals recorded at the scalp. A combination of filter bank common spatial filter for features extraction and support vector machine for classification was used for both pairwise binary classifications and four-class classification of MI tasks. The study provides three major outcomes: 1) the CSD method performs better than CR, providing a significant improvement of 3.02% and 5.59% across six binary classification tasks and four-class classification task, respectively; 2) the combination of a greater number of channels at the pre-processing stage as compared with the feature extraction stage yields better classification accuracies for all the Laplacian methods; and 3) the efficiency of all the surface Laplacian methods reduced significantly in the case of a fewer number of channels considered during the pre-processing.

A novel end‐to‐end deep learning scheme for classifying multi‐class motor imagery electroencephalography signals

“…The large number of published articles on MI–BCI tasks using EEG signals highlights the importance of the applicability of EEG signals in the BCI domain (Hwang et al, ; Ortiz‐Rosario and Adeli, ). Most EEG‐based BCI systems usually use a structured approach that includes three main steps: (a) preprocessing that may contain three suboperations of noise removal (NR; Çınar, S., and Acır, ; Mutanen et al, ), channel selection (CS; Ghaemi et al, ; Rathee et al, ), and data augmentation (Kalunga et al, 2015; Krell et al, 2018); (b) feature construction, that is choosing appropriate properties of signals, consists of two suboperations of feature extraction (Hsu, ; Zhang et al, ; Aghaei et al, ; Cai et al, ), and feature selection (Zhang et al, ; Lin et al, ; Ma et al, ); and (c) classification that is performed using an appropriate classifier such as support vector machine (Khedher et al, ; Dai and Cao, ; Direito et al, ), probabilistic neural networks (Adeli and Panakkat, ; Sankari and Adeli, ), enhanced probabilistic neural network (Ahmadlou and Adeli, ; Hirschauer et al, ; Fernandes et al, ), competitive probabilistic neural network (Zeinali and Story, ), the recently developed neural dynamics classification algorithm (Rafiei and Adeli, ), or a combination or ensemble of classifiers (Oliveira‐Santos et al ; Reyes et al ). It should be noted that the necessity of each aforementioned suboperation is usually determined by a BCI expert, which is not convenient in practice.…”

Online Covariate Shift Detection-Based Adaptive Brain–Computer Interface to Trigger Hand Exoskeleton Feedback for Neuro-Rehabilitation

“…The strong correlation between EEG signals and mental tasks has led to many user centric applications such as virtual spellers for the communication [7], functional electrical stimulation (FES) based neuro-prosthesis for tetraplegics [4], hand exoskeleton control [8], [9], [10], [11] and telepresence for personal assistance [12]. In spite of the seemingly bright prospect of the BCI technology, there are some practical challenges regarding the robustness, accuracy, and information transfer rate (ITR) of such systems [13], [14], [15]. The non-stationary nature of neurophysiological signals and dynamics of brain activity make the EEG-based BCI, a dynamically varying system, and thus improving its learning performance is a challenging task [16].…”

Covariate shift estimation based adaptive ensemble learning for handling non-stationarity in motor imagery related EEG-based brain-computer interface

“…To enhance the performance of MI related BCI systems, various signal processing methods have been proposed to extract effective features in the temporal and spatial domains that can characterise the non-stationarity in EEG signals. For example, in the temporal domain, band-power and band-pass based filtering methods are commonly used [15], whereas in the spatial domain, common averaging, current source density [30], and CSP-based features have been examined for the detection of MI related responses [22,31].…”