Single-trial Motor Imagery EEG Research Articles

Recognizing single-trial motor imagery EEG based on interpretable clustering method

This work explores an approach for single-trial motor imagery (MI) electroencephalography (EEG) classification in interpretable clustering. The tensor structured EEG data under Mu rhythm is first processed by Common Spatial Subspace Decomposition (CSSD) to obtain the multi-dimensional CSSD-mapped EEG. In the dimensional feature reduction, Fisher’s ratio is used as the cost function to automatically find the optimal projection plane with two feature vectors of CSSD-mapped EEG corresponding to the largest Fisher’s ratio. Then, Discriminative Rectangle Mixture Model (DRMM) that gives a rectangular decision rule is used to identify optimal feature vectors to realize single-trial motor imagery EEG classification in an interpretable way. This innovative data analysis model generates reasonable cluster results driven by optimal feature data. The probability density distribution functions of EEG data in two classes can effectively explain the reliability of the rectangular decision rule given by the DRMM. The proposed method has been validated using the areas under the receiver operating characteristic curve (AUC) and cluster quality evaluation metrics. Experimental results demonstrate its performance is comparable to existing clustering and gives interpretable clustering results when detecting the motor intention involving EEG signals. This paper provides a novelty method based on interpretable clustering for single-trial MI EEG classification. And it may promote the development of BCI application.

A brain-computer interface system for smart home control based on single trial motor imagery EEG

In recent years, researches on brain signal recognition and brain-computer interface control have made great progress. By analysing electroencephalogram (EEG), a specific brain activity can be detected and the signal can be used to control smart devices and help people to complete difficult and complicated tasks, especially for people with disabilities. This paper presents the design and implementation of a novel brain-computer interface system for smart home control using single trial motor imagery EEG. The system adopts STM32 Microcontroller Unit (MCU) and ThinkGear Asic Module (TGAM) to realise the acquisition and recognition of EEG signals. It can transfer the signals to portable devices through Bluetooth modules. Three main EEG features including Alpha, Beta, and Gamma waves are discussed. It is tested during simple actions such as blinking in various situations. The experimental results show that the implemented system is suitable for extracting specific EEG signals to control smart home devices.

A brain-computer interface system for smart home control based on single trial motor imagery EEG

A brain-computer interface system for smart home control based on single trial motor imagery EEG

Classification of EEG-based single-trial motor imagery tasks using a B-CSP method for BCI

Classifying single-trial electroencephalogram (EEG) based motor imagery (MI) tasks is extensively used to control brain-computer interface (BCI) applications, as a communication bridge between humans and computers. However, the low signal-to-noise ratio and individual differences of EEG can affect the classification results negatively. In this paper, we propose an improved common spatial pattern (B-CSP) method to extract features for alleviating these adverse effects. First, for different subjects, the method of Bhattacharyya distance is used to select the optimal frequency band of each electrode including strong event-related desynchronization (ERD) and event-related synchronization (ERS) patterns; then the signals of the optimal frequency band are decomposed into spatial patterns, and the features that can describe the maximum differences of two classes of MI are extracted from the EEG data. The proposed method is applied to the public data set and experimental data set to extract features which are input into a back propagation neural network (BPNN) classifier to classify single-trial MI EEG. Another two conventional feature extraction methods, original common spatial pattern (CSP) and autoregressive (AR), are used for comparison. An improved classification performance for both data sets (public data set: 91.25%±1.77% for left hand vs. foot and 84.50%±5.42% for left hand vs. right hand; experimental data set: 90.43%±4.26% for left hand vs. foot) verifies the advantages of the B-CSP method over conventional methods. The results demonstrate that our proposed B-CSP method can classify EEG-based MI tasks effectively, and this study provides practical and theoretical approaches to BCI applications.

Classification of single-trial motor imagery EEG by complexity regularization

Brain computer interface based on electroencephalogram is a popular way to enable communication between brain and output devices helping elderly and disabled people and in rehabilitation. In practice, the effectiveness of brain computer interface has a strong relationship with the classification accuracy of single trials. Common spatial pattern is believed to be an effective algorithm for classifying the single-trial brain signal. Since it is based on the characteristics of a broad frequency band which is manually selected and not individual variability, it is sensitive to noise and individual variability. In this paper, the common spatial pattern was extended in order to improve classification accuracies and to mitigate these influences. The channel-specific complexity weights of characteristic on montage were derived and added to improve the effects of the relevant function area and the separability between classes. The proposed method was evaluated using two public datasets, and achieved an average accuracy of 18.4% higher than conventional common spatial pattern, and the performance of the proposed method over conventional common spatial pattern was significant (p < 0.05). It indicates that the proposed method extracts subject-specific characteristics and outperforms the conventional common spatial pattern in single-trial EEG classification.

Single-trial EEG classification of motor imagery using deep convolutional neural networks

Electroencephalogram (EEG) signal recorded during motor imagery (MI) has been widely applied in non-invasive brain–computer interface (BCI) as a communication approach. In this paper, we propose a new method based on the deep convolutional neural network (CNN) to perform feature extraction and classification for single-trial MI EEG. Firstly, based on the spatio-temporal characteristics of EEG, a 5-layer CNN model is built to classify MI tasks (left hand and right hand movement); then the CNN model is applied in the experimental data set collected from subjects, and compared with other three conventional classification methods (power+SVM, CSP+SVM and AR+SVM). The results demonstrate that CNN can further improve classification performance: the average accuracy using CNN (86.41±0.77%) is 9.24%, 3.80% and 5.16% higher than those using power+SVM, CSP+SVM and AR+SVM, respectively. The present study shows that the proposed method is effective to classify MI, and provides a practical method by non-invasive EEG signal in BCI applications.

Channel selection by Rayleigh coefficient maximization based genetic algorithm for classifying single-trial motor imagery EEG

While common spatial pattern may be the most widely used feature for discriminating motor imagery based EEG signals, Rayleigh coefficient maximization enable us to have one more effective. However, such a feature is often deteriorated by redundant electrode channels which may result in low classification accuracy, extra subsequent computational load and difficulty in understanding which part of the brain relates to classification-relevant activity. In this paper, we present a channel selection method to deal with these problems, in which an improved genetic algorithm based on the Rayleigh coefficient feature is conducted to determine the optimal subset of channels. Experiment results on two motor imagery EEG datasets verify that our method is effective in channel selection for classifying motor imagery EEG signals.

Classification of single trial motor imagery EEG recordings with subject adapted non-dyadic arbitrary time–frequency tilings

We describe a new technique for the classification of motor imagery electroencephalogram (EEG) recordings in a brain computer interface (BCI) task. The technique is based on an adaptive time–frequency analysis of EEG signals computed using local discriminant bases (LDB) derived from local cosine packets (LCP). In an offline step, the EEG data obtained from the C3/C4 electrode locations of the standard 10/20 system is adaptively segmented in time, over a non-dyadic grid by maximizing the probabilistic distances between expansion coefficients corresponding to left and right hand movement imagery. This is followed by a frequency domain clustering procedure in each adapted time segment to maximize the discrimination power of the resulting time–frequency features. Then, the most discriminant features from the resulting arbitrarily segmented time-frequency plane are sorted. A principal component analysis (PCA) step is applied to reduce the dimensionality of the feature space. This reduced feature set is finally fed to a linear discriminant for classification. The online step simply computes the reduced dimensionality features determined by the offline step and feeds them to the linear discriminant. We provide experimental data to show that the method can adapt to physio-anatomical differences, subject-specific and hemisphere-specific motor imagery patterns. The algorithm was applied to all nine subjects of the BCI Competition 2002. The classification performance of the proposed algorithm varied between 70% and 92.6% across subjects using just two electrodes. The average classification accuracy was 80.6%. For comparison, we also implemented an adaptive autoregressive model based classification procedure that achieved an average error rate of 76.3% on the same subjects, and higher error rates than the proposed approach on each individual subject.