Abstract:Studies have shown that a motor imagery electro encephalogram (EEG)-based brain-computer interface (BCI) system can be used as a rehabilitation tool for stroke patients. Efficient classification of EEG from stroke patients is fundamental in the BCI-based stroke rehabilitation systems. One of the most successful algorithms for EEG classification is the common spatial patterns (CSP). However, studies have reported that the performance of CSP heavily relies on its operational frequency band and channels configuration. To the best of our knowledge, there is no agreed upon clinical conclusion about motor imagery patterns of stroke patients. In this case, it is not available to obtain the active channels and frequency bands related to brain activities of stroke patients beforehand. Hence, for using the CSP algorithm, we usually set a relatively broad frequency range and channels, or try to find subject-related frequency bands and channels. To address this problem, we propose an adaptive boosting algorithm to perform autonomous selection of key channels and frequency band. In the proposed method, the spatial-spectral configurations are divided into multiple preconditions, and a new heuristic supervisor of stochastic gradient boost strategy is utilized to train weak classifiers under these preconditions. Extensive experiment comparisons have been performed on three datasets including two benchmark datasets from the famous BCI competition III and BCI competition IV as well as one self-acquired dataset from stroke patients. Results show that our algorithm yields relatively higher classification accuracies compared with seven state-of-the-art approaches. In addition, the spatial patterns (spatial weights) and spectral patterns (bandpass filters) determined by the algorithm can also be used for further analysis of the data, e.g., for brain source localization and physiological knowledge exploration.

Gaussian Mixture Modeling In Stroke Patients' Rehabilitation Eeg Data Analysis

An Iterative Method for Classifying Stroke Subjects’ Motor Imagery EEG Data in the BCI-FES Rehabilitation Training System

Spatial-Spectral Boosting Analysis for Stroke Patients' Motor Imagery EEG in Rehabilitation Training

A frequency boosting method for motor imagery EEG classification in BCI-FES rehabilitation training system

EEG Classification by Filter Band Component Regularized Common Spatial Pattern for Motor Imagery.

A Boosting-Based Spatial-Spectral Model for Stroke Patients’ EEG Analysis in Rehabilitation Training

An optimized GMM algorithm and its application in single-trial motor imagination recognition

A tensor-based scheme for stroke patients' motor imagery EEG analysis in BCI-FES rehabilitation training.

Multi-Task Heterogeneous Ensemble Learning-Based Cross-Subject EEG Classification Under Stroke Patients

Exploring Motor Imagery Eeg Patterns for Stroke Patients with Deep Neural Networks

Enhancing Motor Imagery Classification with Residual Graph Convolutional Networks and Multi-Feature Fusion

Time-Frequency-Space EEG Decoding Model Based on Dense Graph Convolutional Network for Stroke

Motor imagery electroencephalogram classification algorithm based on joint features in the spatial and frequency domains and instance transfer

Single-Trial Discrimination Of Eeg Signals For Stroke Patients: A General Multi-Way Analysis

Deep Convolution Generative Adversarial Network-Based Electroencephalogram Data Augmentation for Post-Stroke Rehabilitation with Motor Imagery

Scale-Dependent Signal Identification in Low-Dimensional Subspace: Motor Imagery Task Classification

Self-Supervised EEG Representation Learning with Contrastive Predictive Coding for Post-Stroke Patients

Gaussian mixture model based on genetic algorithm for brain-computer interface

Motor Imagery Recognition Based on GMM-JCSFE Model

Stroke Motor Imagery Recognition based on EEG

EEG Neuro-markers to Enhance BCI-based Stroke Patients Rehabilitation