Brain-computer Interface System Research Articles

Decoding continuous motion trajectories of upper limb from EEG signals based on feature selection and nonlinear methods.

Brain-computer interface (BCI) system has emerged as a promising technology that provides direct communication and control between the human brain and external devices. Among the various applications of BCI, limb motion decoding has gained significant attention due to its potential for patients with motor impairment to regain independence and improve their quality of life. However, the reconstruction of continuous motion trajectories in BCI systems based on electroencephalography (EEG) signals remains a challenge in practical life. This study investigates the feasibility of applying feature selection and nonlinear regression for decoding motion trajectory from EEG. We propose to fix the time window, select the optimal feature set, and reconstruct the motion trajectory of motor execution tasks using polynomial regression. The proposed approach is validated on a public dataset consisting of EEG and hand position data recorded from 15 subjects. Several methods including ridge regression and multiple linear regression are employed as comparisons. The cross-validation results show that the proposed reconstructed method has the highest correlation with actual motion trajectories, with an average value of 0.511±0.019 (p<0.05). This finding demonstrates the great potential of our approach for real-world motor kinematics BCI applications.

Design and implementation of a scalable and high-throughput EEG acquisition and analysis system

Recent advances in neuroscience, neuromorphic intelligence, and brain–computer interface (BCI) technologies have created a need for fast, efficient, and convenient electroencephalogram (EEG) data acquisition systems. However, the existing equipment was limited in its flexibility, restricting non-invasive studies to research or medical settings. To address this issue, low-cost, compact EEG acquisition devices have been developed, allowing for frequent and flexible brain data acquisition in various scenarios. This paper introduces a scalable and high-throughput EEG signal acquisition and analysis system based on field-programmable gate array (FPGA) technology. The proposed system offers electrode scalability, on-chip computing, and optional wireless functionality extension. These features are achieved through the design of a highly scalable underlying EEG acquisition module and an FPGA central module that enables software-defined high-throughput expansion and high-speed data exchange between software and hardware. The paper presents two implementation cases that demonstrate the potential of the proposed system. The first case introduces a wearable wireless EEG system, enabling the deployment of effective and user-friendly steady-state visual evoked potential (SSVEP)-BCI applications in consumer-grade scenarios. The second case integrates an FPGA central module with multiple basic EEG acquisition modules to construct a high-throughput BCI system for cost-effective and real-time EEG data acquisition and processing. This configuration allows for flexible deployment in research and clinical applications, including attention index, SSVEP, motor imagery (MI), and emotion recognition. This combination further demonstrates the potential of scalable EEG systems and emphasizes the need for further integration or chipization. These implementations validate the feasibility of compact and efficient EEG devices and highlight the promising applications of scalable BCI system in various fields.

Developing Innovative Feature Extraction Techniques from the Emotion Recognition Field on Motor Imagery Using Brain–Computer Interface EEG Signals

Research on brain–computer interfaces (BCIs) advances the way scientists understand how the human brain functions. The BCI system, which is based on the use of electroencephalography (EEG) signals to detect motor imagery (MI) tasks, enables opportunities for various applications in stroke rehabilitation, neuroprosthetic devices, and communication tools. BCIs can also be used in emotion recognition (ER) research to depict the sophistication of human emotions by improving mental health monitoring, human–computer interactions, and neuromarketing. To address the low accuracy of MI-BCI, which is a key issue faced by researchers, this study employs a new approach that has been proven to have the potential to enhance motor imagery classification accuracy. The basic idea behind the approach is to apply feature extraction methods from the field of emotion recognition to the field of motor imagery. Six feature sets and four classifiers were explored using four MI classes (left and right hands, both feet, and tongue) from the BCI Competition IV 2a dataset. Statistical, wavelet analysis, Hjorth parameters, higher-order spectra, fractal dimensions (Katz, Higuchi, and Petrosian), and a five-dimensional combination of all five feature sets were implemented. GSVM, CART, LinearSVM, and SVM with polynomial kernel classifiers were considered. Our findings show that 3D fractal dimensions predominantly outperform all other feature sets, specifically during LinearSVM classification, accomplishing nearly 79.1% mean accuracy, superior to the state-of-the-art results obtained from the referenced MI paper, where CSP reached 73.7% and Riemannian methods reached 75.5%. It even performs as well as the latest TWSB method, which also reached approximately 79.1%. These outcomes emphasize that the new hybrid approach in the motor imagery/emotion recognition field improves classification accuracy when applied to motor imagery EEG signals, thus enhancing MI-BCI performance.

A Hybrid RNN-CNN based Motor Imagery Tasks Classification Approach Using MEG Brain Signals for BCI Applications

&lt;p dir="ltr"&gt;&lt;span&gt;Magnetoencephalography (MEG) has become a pioneering technology in Brain-Computer Interfaces (BCIs) for neurorehabilitation, which significantly improves communication and motor rehabilitation for people with neurological conditions, especially stroke survivors. MEG-based BCIs allow users to regain control over their motor and cognitive functions by providing precise temporal and spatial resolution for detecting neural activity. The MEG has several benefits including reduced distortion from the skull and scalp, its non-invasive nature, and the ability to capture deep brain activity, all of which enhance the effectiveness of BCI systems. This study focuses on classifying motor and cognitive tasks using various models, including Recurrent Neural Networks (RNNs), one-dimensional Convolutional Neural Networks (1DCNNs), and a hybrid approach. The effectiveness of the models in classification tasks was evaluated using various metrics such as accuracy, recall, precision, and F1-score. Significantly, the hybrid model exhibited enhanced performance relative to the other models, achieving marked improvements in both classification accuracy and robustness. These results underscore the promise of MEG BCI technology for neurorehabilitation and stress the need for the development of sophisticated classification models to support the recovery of motor and cognitive abilities in people with neurological conditions. This study adds to the expanding research focused on enhancing the quality of life for affected individuals by utilizing innovative BCI solutions, leveraging the unique capabilities of MEG technology to enable more effective neurorehabilitation interventions.&lt;/span&gt;&lt;/p&gt;

Multi-layer ear-scalp distillation framework for ear-EEG classification enhancement

Ear-electroencephalography (ear-EEG) holds significant promise as a practical tool in brain-computer interfaces (BCIs) due to its enhanced unobtrusiveness, comfort, and mobility compared to traditional steady-state visual evoked potential (SSVEP)-based BCI systems. However, achieving accurate SSVEP classification with ear-EEG remains a major challenge due to the significant attenuation and distortion of the signal amplitude. Our aim is to enhance the classification performance of SSVEP using ear-EEG and to increase its practical application value. To address this challenge, we focus on enhancing ear-EEG feature representations by training the model to learn features similar to those of scalp-EEG. We introduce a novel framework, termed multi-layer ear-scalp distillation (MESD), designed to optimize SSVEP target classification in ear-EEG data. This framework combines signals from the scalp to obtain multi-layer distilled knowledge through the cooperation of mid-layer feature distillation and output layer response distillation.Mainresults.We improve the classification of the initial 1 s data and achieved a maximum classification accuracy of 81.1%. We evaluate the proposed MESD framework through single-session, cross-session, and cross-subject transfer decoding, comparing it with baseline methods. The results demonstrate that the proposed framework achieves the best classification results in all experiments. Our study enhances the classification accuracy of SSVEP based on ear-EEG within a short time window. These results offer insights for the application of ear-EEG brain-computer interfaces in tasks such as auxiliary control and rehabilitation training in future endeavors.

An Empirical Model-Based Algorithm for Removing Motion-Caused Artifacts in Motor Imagery EEG Data for Classification Using an Optimized CNN Model.

Electroencephalography (EEG) is a non-invasive technique with high temporal resolution and cost-effective, portable, and easy-to-use features. Motor imagery EEG (MI-EEG) data classification is one of the key applications within brain-computer interface (BCI) systems, utilizing EEG signals from motor imagery tasks. BCI is very useful for people with severe mobility issues like quadriplegics, spinal cord injury patients, stroke patients, etc., giving them the freedom to a certain extent to perform activities without the need for a caretaker, like driving a wheelchair. However, motion artifacts can significantly affect the quality of EEG recordings. The conventional EEG enhancement algorithms are effective in removing ocular and muscle artifacts for a stationary subject but not as effective when the subject is in motion, e.g., a wheelchair user. In this research study, we propose an empirical error model-based artifact removal approach for the cross-subject classification of motor imagery (MI) EEG data using a modified CNN-based deep learning algorithm, designed to assist wheelchair users with severe mobility issues. The classification method applies to real tasks with measured EEG data, focusing on accurately interpreting motor imagery signals for practical application. The empirical error model evolved from the inertial sensor-based acceleration data of the subject in motion, the weight of the wheelchair, the weight of the subject, and the surface friction of the terrain under the wheelchair. Three different wheelchairs and five different terrains, including road, brick, concrete, carpet, and marble, are used for artifact data recording. After evaluating and benchmarking the proposed CNN and empirical model, the classification accuracy achieved is 94.04% for distinguishing between four specific classes: left, right, front, and back. This accuracy demonstrates the model's effectiveness compared to other state-of-the-art techniques. The comparative results show that the proposed approach is a potentially effective way to raise the decoding efficiency of motor imagery BCI.

The application of integrating electroencephalograph-based emotion recognition technology into brain–computer interface systems for the treatment of depression: a narrative review

Dysregulation of the prefrontal cortex, amygdala, and hippocampus, along with alterations in P300 amplitude and abnormalities in the theta and beta bands, has been closely linked to the onset and pathophysiology of depression. Consequently, integrating electroencephalograph-based emotion recognition technology into brain‒computer interface systems offers the potential for real-time identification and modulation of emotional states through continuous interaction between the brain‒computer interface system and brain activity. This closed-loop system could precisely control neural stimulation in brain regions associated with emotional disorders, potentially alleviating the distressing memories of traumatic events. Although the efficacy of the brain‒computer interface in treating depression still requires validation through extensive clinical trials, its inherent real-time feedback and adaptive capabilities present a promising avenue for depression therapy. This review aims to explore the neuroanatomical mechanisms and neural activity patterns associated with depression and evaluate the potential of brain‒computer interface technology as a treatment modality. The objectives include summarizing key brain regions and neural networks involved in depression, analyzing their activity patterns, and assessing the impact of brain‒computer interface technology on these regions to provide theoretical support for future clinical trials. Significant functional abnormalities have been identified in the prefrontal cortex, amygdala, and hippocampus of patients with depression. The gray matter density, functional connectivity, and neural activity in these regions are closely associated with the severity of depressive symptoms. Common features in patients with depression include a reduced P300 amplitude and increased θ and α current density. Brain‒computer interface technology has demonstrated potential in modulating these abnormal neural activities, particularly in emotion recognition and regulation. When combined with techniques such as repetitive transcranial magnetic stimulation and deep brain stimulation, brain‒computer interface may provide effective interventions for managing emotional states in patients with depression. This review confirms the association between depression and functional abnormalities in specific brain regions and suggests that brain‒computer interface technology offers promising therapeutic potential by modulating abnormal neural activity. Brain‒computer interface could represent a novel treatment approach for depression. Future research should focus on validating the practical applications, efficacy, and safety of brain‒computer interface in treating depression.

Exoskeleton Robot Based on Brain-Computer Interface

Brain-computer interface (BCI) technology represents a means of facilitating human-computer interaction. One of the most widely accepted paradigms of brain-computer interface is motor imagery, which enables the recognition of electroencephalogram (EEG) signals generated in a specific brain region by imagining the movement of a limb. Following the acquisition, preprocessing, feature processing, and signal classification of the EEG signals, the complex signals are accurately recognized. Therefore, by creating a control system that translates the recognized EEG signals into movement commands for the robot and transmits them to the robot, it is possible to control the robot's movements by motor imagery. The convolutional neural network is the most popular signal processing algorithm due to its high EEG recognition accuracy, excellent performance in feature extraction, and superior performance in end-to-end learning. The convolutional neural network is an optimal method for signal processing in robot control. This makes CNN an optimal choice for processing EEG signals in robot control, enhancing both the effectiveness and user experience of BCI systems by enabling more intuitive and responsive interactions with robotic devices.

A reinforcement learning based software simulator for motor brain-computer interfaces.

Intracortical motor brain-computer interfaces (BCIs) are expensive and time-consuming to design because accurate evaluation traditionally requires real-time experiments. In a BCI system, a user interacts with an imperfect decoder and continuously changes motor commands in response to unexpected decoded movements. This "closed-loop" nature of BCI leads to emergent interactions between the user and decoder that are challenging to model. The gold standard for BCI evaluation is therefore real-time experiments, which significantly limits the speed and community of BCI research. We present a new BCI simulator that enables researchers to accurately and quickly design BCIs for cursor control entirely in software. Our simulator replaces the BCI user with a deep reinforcement learning (RL) agent that interacts with a simulated BCI system and learns to optimally control it. We demonstrate that our simulator is accurate and versatile, reproducing the published results of three distinct types of BCI decoders: (1) a state-of-the-art linear decoder (FIT-KF), (2) a "two-stage" BCI decoder requiring closed-loop decoder adaptation (ReFIT-KF), and (3) a nonlinear recurrent neural network decoder (FORCE).

A Review on Machine Learning Based on EEG and ECoG Signal for Brain Computer Interface

Abstract. Brain-computer interfaces (BCIs) are devices utilized for the detection and recognition of brain activity. With advancements in machine learning techniques, BCI projects have witnessed an increasing application of these techniques, leading to higher performance. BCI also finds utility in medical applications by providing efficient solutions for individuals with neurological disorders or disabilities. The research focus on convenient and non-invasive electroencephalography (EEG) and high signal-to-noise ratio electrocorticography (ECoG) stems from their respective advantages. This paper presents a comprehensive overview of BCI principles, imaging techniques such as EEG and ECoG along with their advantages and disadvantages, as well as discusses the utilization of various machine learning techniques in EEG and ECoG-based BCI system, including classification and regression models. Furthermore, this paper summarizes the challenges faced by EEG and ECoG-based BCI techniques while also discussing future directions such as artifacting removal technique for EEG-based BCIs and closed-loop feedback technique based on ECoGBCI.

Classification of motor imagery EEG with ensemble RNCA model

Motor Imagery (MI) based brain-computer interface (BCI) systems are used for regaining the motor functions of neurophysiologically affected persons. But the performance of MI tasks is degraded due to the presence of redundant EEG channels. Hence, a novel ensemble regulated neighborhood component analysis (ERNCA) method provides a perfect identification of neural region that stimulate motor movements. Domains of statistical, frequency, spatial and transform-based features narrowed down the misclassification rate. The gradient boosting method selects the relevant features thereby reduces the computational complexity. Finally, Bayesian optimized ensemble classifier finetuned the classification accuracies of 97.22 % and 91.62 % for Datasets IIIa and IVa respectively. This approach is further strengthened by analyzing real-time data with the accuracy of 93.75 %. This method qualifies out of four benchmark methods with significant percent of improvement in accuracy for these three datasets. As per the spatial distribution of refined EEG channels, majority of the brain's motor functions concentrates on frontal and central cortex regions of brain.

Population-based evolutionary search for joint hyperparameter and architecture optimization in brain-computer interface

In recent years, deep learning (DL)-based models have become the de facto standard for motor imagery brain-computer interface (MI-BCI) systems due to their notable performance. However, these models often require extensive hyperparameter optimization process to achieve optimal results. To tackle this challenge, recent studies have proposed various methods to automate this process. Despite promising results, these methods overlook the architecture elements, which are crucial factors for MI-BCI system performance and are highly intertwined with hyperparameter settings. To overcome this limitation, we propose a joint optimization framework that uses a population-based evolutionary search to optimize both hyperparameters and architectures. Our framework adopts a two-stage optimization approach that alternates between hyperparameter and architecture optimization to effectively manage the complexity of the joint search process. Furthermore, we introduce a novel ensemble method that leverages diverse promising configurations to enhance generalization and robustness. Evaluations on two public MI-BCI datasets show that our framework consistently outperforms competing methods across a range of backbone models, demonstrating its effectiveness and versatility.

MSMGE-CNN: a multi-scale multi-graph embedding convolutional neural network for motor related EEG decoding

Abstract Deep learning (DL) technique has been widely used for decoding motor related electroencephalography (EEG) signals, which has considerably driven the development of motor related brain-computer interfaces (BCIs). However, traditional convolutional neural networks (CNNs) cannot fully represent spatial topology information and dynamic temporal characteristics of multi-channel EEG signals, resulting in limited decoding accuracy. To address such challenges, a novel multi-scale multi-graph embedding convolutional neural network (MSMGE-CNN) is proposed in this study. The proposed MSMGE-CNN contains two crucial components: multi-scale time convolution and multi-graph embedding. Specifically, we design a multi-branch CNN architecture with mixed-scale time convolutions based on EEGNet to sufficiently extract robust time domain features. Afterward, we embed multi-graph information obtained based on physical distance proximity and functional connectivity of multi-channel EEG signals into the time-domain features to capture rich spatial topological dependencies via multi-graph convolution operation. We extensively evaluated the proposed method on three benchmark EEG datasets commonly used for motor imagery/execution (MI/ME) classification and obtained accuracies of 79.59 % (BCICIV-2a Dataset), 69.77 % (OpenBMI Dataset) and 96.34 % (High Gamma Dataset), respectively. These results powerfully demonstrate that MSMGE-CNN outperforms several state-of-the-art algorithms. In addition, we further conducted a series of ablation experiments to validate the rationality of our network architecture. Overall, the proposed MSMGE-CNN method dramatically improves the accuracy and robustness of MI/ME-EEG decoding, which can effectively enhance the performance of motor related BCI system.&amp;#xD;

Enhancing detection of SSVEP-based BCIs via a novel temporally local canonical correlation analysis

BackgroundIn recent years, spatial filter-based frequency recognition methods have become popular in steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems. However, these methods are ineffective in suppressing local noise, and they rely on the length of the data. In practical applications, enhancing recognition performance with short data windows is a significant challenge for the BCI systems. New methodWith extracting temporal information and eliminating local noise, a temporally local canonical correlation analysis based on training data-driven (TI-tdCCA) method is proposed to enhance the recognition performance of SSVEPs. Based on a novel framework, the filters are derived by incorporating the Laplacian matrix through the use of TI-CCA between the concatenated training data and individual templates. The target frequency is subsequently determined by applying the appropriate spatial filters and Laplacian matrix. ResultsThe experimental results on two datasets, consisting of 40 classes and recording from 35 and 70 subjects respectively, demonstrate that the proposed method consistently outperforms the eight competing methods in the majority of cases. The proposed method is simultaneously evaluated by an extended version that incorporates artificial reference signals. The extended method demonstrates a significant improvement over the proposed method. Specifically, with a time window of 0.7 s, the average recognition accuracy of the subjects increases by 10.71 % on the Benchmark dataset and by 6.98 % on the BETA dataset, respectively. Comparison with existing methodsOur extended method outperforms the state-of-the-art methods by at least 3 %, and it effectively suppresses local noise and maintains excellent scalability. Conclusions for research articlesThe proposed method can effectively combine spatial and temporal filters to improve the recognition performance of SSVEPs.

Classification of hand movements from EEG using a FusionNet based LSTM network

Objective. Accurate classification of electroencephalogram (EEG) signals is crucial for advancing brain-computer interface (BCI) technology. However, current methods face significant challenges in classifying hand movement EEG signals, including effective spatial feature extraction, capturing temporal dependencies, and representing underlying signal dynamics.Approach. This paper introduces a novel multi-model fusion approach, FusionNet-Long Short-Term Memory (LSTM), designed to address these issues. Specifically, it integrates Convolutional Neural Networks for spatial feature extraction, Gated Recurrent Units and LSTM networks for capturing temporal dependencies, and Autoregressive (AR) models for representing signal dynamics.Main results. Compared to single models and state-of-the-art methods, this fusion approach demonstrates substantial improvements in classification accuracy. Experimental results show that the proposed model achieves an accuracy of 87.1% in cross-subject data classification and 99.1% in within-subject data classification. Additionally, Gradient Boosting Trees were employed to evaluate the significance of various EEG features to the model.Significance. This study highlights the advantages of integrating multiple models and introduces a superior classification model, which is pivotal for the advancement of BCI systems.

Spatialspectral-Backdoor: Realizing backdoor attack for deep neural networks in brain–computer interface via EEG characteristics

In recent years, electroencephalogram (EEG) based on the brain–computer interface (BCI) systems have become increasingly advanced, with researcher using deep neural networks as tools to enhance performance. BCI systems heavily rely on EEG signals for effective human–computer interactions, and deep neural networks show excellent performance in processing and classifying these signals. Nevertheless, the vulnerability to backdoor attack is still a major problem. Backdoor attack is the injection of specially designed triggers into the model training process, which can lead to significant security issues. Therefore, in order to simulate the negative impact of backdoor attack and bridge the research gap in the field of BCI, this paper proposes a new backdoor attack method to call researcher attention to the security issues of BCI. In this paper, Spatialspectral-Backdoor is proposed to effectively attack the BCI system. The method is carefully designed to target the spectral active backdoor attack of the BCI system and includes a multi-channel preference method to select the electrode channels sensitive to the target task. Ultimately, the effectiveness of the comparison and ablation experiments is validated on the publicly available BCI competition datasets. The results show that the average attack success rate and clean sample accuracy of Spatialspectral-Backdoor in the BCI scenario are 97.12% and 85.16%, respectively, compared with other backdoor attack methods. Furthermore, by observing the infection ratio of backdoor triggers and visualization of the feature space, the proposed Spatialspectral-Backdoor outperforms other backdoor attack methods.

A study of external machine control pathways based on brain-computer interfaces

Abstract. This research analyses three fundamental paradigms in brain-computer interface (BCI) systems: motor imagery (MI), sensorimotor rhythms (SMRs), and P300-based paradigms, focusing on their methodology and applications. Motor imagery-based brain-computer interface (MI-BCI) entails mentally practicing motor activities while adjusting electroencephalography (EEG) patterns and utilising multimodal stimulation to enhance the accuracy of classification. SMR-BCI employs precise EEG band oscillations to operate prosthetics and virtual environments. Sensory threshold neuromuscular electrical stimulation is used to improve performance. The P300 paradigm utilizes event-related potentials to control devices, and includes advancements such as the omitted elicited paradigm to enhance efficiency. In addition, the study examines control strategies for BCI systems, such as PID, MPC, and shared control approaches. By utilizing MATLAB as a tool, one may ascertain the distinction between different control strategies, highlighting their respective contributions to improving system responsiveness and stability.

A Novel Filter Bank and Fourier Transform Convolutional Neural Network for SSVEP Classification

Abstract. Classifying Steady-State Visual Evoked Potentials (SSVEPs) is crucial for enhancing the performance of Brain-Computer Interface (BCI) systems. This study introduces a new method for SSVEP classification within BCI frameworks, called the Filter Bank and Fourier Transform Convolutional Neural Network (FBFCNN). The FBFCNN model effectively extracts key harmonic features by first segmenting SSVEP signals into multiple sub-bands through a filter bank. Subsequently, the Fast Fourier Transform (FFT) is applied to convert these signals from the time domain into the frequency domain. The generated spectrum's real and imaginary components are then fed into a convolutional neural network (CNN). This method improves SSVEP feature extraction and increases classification accuracy by fusing the advantages of CNNs with filter banks. Experimental results, derived from publicly available SSVEP datasets, show that the FBFCNN model surpasses traditional techniques in both accuracy and Information Transfer Rate (ITR). Offering a reliable and efficient solution for real-time brain signal decoding, the FBFCNN model represents a significant advancement in SSVEP-based BCI systems.

Advancements in EEG-based BCI for cognitive decline diagnosis using deep learning and Transformers

Abstract. Deep learning and Transformer models have revolutionized medical diagnostics, particularly in the advanced analysis of large datasets. Given the individual variability in cognitive decline status, the accuracy of diagnosing elderly cognitive decline diseases is crucial. Since it allows for the initiation of treatment and interventions at the earliest possible stage, potentially slowing the progression of the disease. In this paper, the author reviews and summarizes studies on the use of Transformers and deep learning in Electroencephalography (EEG) signal decoding, which is the most prevalent signal-detecting method in Brain-Computer Interface (BCI) systems for elderly cognitive decline. The review indicates that deep learning and transformer models outperform traditional methods in classifying cognitive decline, offering effective methods and improved accuracy by extracting and analyzing in-depth features from EEG data. Differently, deep learning models are proficient in single modality learning, but transformer models offer a unified approach to process and integrate information from diverse data sources and, with the increase in data volume, Transformers have the potential to further improve the accuracy of diagnosis.

Incremental Classification for High-Dimensional EEG Manifold Representation Using Bidirectional Dimensionality Reduction and Prototype Learning.

In brain-computer interface (BCI) systems, symmetric positive definite (SPD) manifold within Riemannian space has been frequently utilized to extract spatial features from electroencephalogram (EEG) signals. However, the intrinsic high dimensionality of SPD matrices introduces too much computational burden to hinder the real-time applications of such BCI, especially in handling dynamic tasks, like incremental learning. Directly reducing the dimensionality of SPD matrices with conventional dimensionality reduction (DR) methods will alter the fundamental properties of SPD matrices. Moreover, current DR methods for incremental learning always necessitate retaining old data to update their representations under new mapping. To this end, a bidirectional two-dimensional principal component analysis for SPD manifold (B2DPCA-SPD) is proposed to reduce the dimensionality of SPD matrices, in such way that the reduced matrices remain on SPD manifold. Afterwards, the B2DPCA-SPD is extended to adapt to incremental learning tasks without saving old data. The incremental B2DPCA-SPD can be seamlessly integrated with the matrix-formed growing neural gas network (MF-GNG) to achieve an incremental EEG classification, where the new low-dimensional representations of the prototypes in old classifiers can be easily recalculated with the updated projection matrix. Extensive experiments are conducted on two public datasets to perform the EEG classification. The results demonstrate that our method significantly reduces computation time by 38.53% and 35.96%, and outperforms conventional methods in classification accuracy by 4.21% to 19.59%.