Filter Bank Common Spatial Pattern Algorithm on BCI Competition IV Datasets 2a and 2b

By Kai Keng Ang, Z. Y. Chin, Chuanchu Wang, Cuntai Guan, Haihong Zhang. Frontiers in Neuroscience, 2012

1,214 citations recorded by the source corpus.

The Common Spatial Pattern (CSP) algorithm is an effective and popular method for classifying 2-class motor imagery electroencephalogram (EEG) data, but its effectiveness depends on the subject-specific frequency band. This paper presents the Filter Bank Common Spatial Pattern (FBCSP) algorithm to optimize the subject-specific frequency band for CSP on Datasets 2a and 2b of the Brain-Computer Interface (BCI) Competition IV. Dataset 2a comprised 4 classes of 22 channels EEG data from 9 subjects, and Dataset 2b comprised 2 classes of 3 bipolar channels EEG data from 9 subjects. Multi-class extensions to FBCSP are also presented to handle the 4-class EEG data in Dataset 2a, namely, Divide-and-Conquer (DC), Pair-Wise (PW), and One-Versus-Rest (OVR) approaches. Two feature selection algorithms are also presented to select discriminative CSP features on Dataset 2b, namely, the Mutual Information-based Best Individual Feature (MIBIF) algorithm, and the Mutual Information-based Rough Set Reduction (MIRSR) algorithm. The single-trial classification accuracies were presented using 10 × 10-fold cross-validations on the training data and session-to-session transfer on the evaluation data from both datasets. Disclosure of the test data labels after the BCI Competition IV showed that the FBCSP algorithm performed relatively the best among the other submitted algorithms and yielded a mean kappa value of 0.569 and 0.600 across all subjects in Datasets 2a and 2b respectively.

Subfields: Cognitive Neuroscience, Human-Computer Interaction, Signal Processing

View Paper: https://www.semanticscholar.org/paper/bc122aa1332985c4b346fc34b5cfdd65c22a08ea

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Human-Computer Interaction

In Signal Processing

Citations in the corpus

LSTM-Based EEG Classification in Motor Imagery Tasks (2018, background): “Experimental results demonstrate that the RBM-based classification framework outperforms some state-of-the-arts, such as filter bank common spatial pattern (FBCSP) in combination with mutual information-based best individual feature (MIBIF) [26] and FBCSP in combination with mutual information-based”
A novel deep learning approach for classification of EEG motor imagery signals (2017, methodology): “In the classification part, many traditional algorithms like support vector machine (SVM) [2, 16], linear discriminant analysis (LDA) [17] and Bayesian classifier [18] have been employed in different studies [8, 15, 19].”
Review of the BCI Competition IV (2012, methodology): “A more detailed explanation of the winning algorithm is given in a separate paper (Ang et al., 2012).”
Physics-Informed Attention Temporal Convolutional Network for EEG-Based Motor Imagery Classification (2023, background): “filter bank common spatial patterns (FBCSP) [3], and its variants achieved the best performance in MI classification.”
Noninvasive neuroimaging enhances continuous neural tracking for robotic device control (2019, methodology): “Spatial filtering has long been used to denoise noninvasive BCI signals (22, 23) and has recently offered promise in detecting increasingly diverse realistic commands (10, 24, 25).”
EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization (2022, methodology): “For example, FBCSP [8], the winner of BCI Competition IV using hand-crafted spatial features; Con-vNet [16] and EEGNet [17], which have shown remarkable results on many EEG datasets with CNN-based end-to-end frameworks; C2CM [28], which inputs the FBCSP features to the CNN model, combining the…”
Learning Temporal Information for Brain-Computer Interface Using Convolutional Neural Networks (2018, methodology): “The filter-bank CSP (FBCSP) algorithm [15], [16] extends the CSP algorithm by considering that not all frequency bands contain discriminative information, and therefore, by passing”
Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain–computer interface (2015, methodology): “For example, a ﬁ lter bank common spatial pattern (FBCSP) method was proposed to extract EEG power features in mu-and beta-frequency bands and signi ﬁ cantly improved the discrimination of different motor imagery states (Ang et al 2012).”
Signal Processing Approaches to Minimize or Suppress Calibration Time in Oscillatory Activity-Based Brain–Computer Interfaces (2015, methodology): “…and classifier optimization from limited training data), several algorithms being available to optimize such bands with CSP and LDA; see, e.g., [75]–[81]; 2) outlier/ artifact removal in small sample size settings, as when little data is available, a single outlier (artefacts being…”
Deep learning techniques for classification of electroencephalogram (EEG) motor imagery (MI) signals: a review (2021, methodology): “Filter bank CSP (FBCSP) [23] is another extended version of the CSP approach that uses the spatial information in EEG channels, as well as the frequency data in MI-EEG signals.”
Macroscopic EEG reveals discriminative low-frequency oscillations in plan-to-grasp visuomotor tasks (2025, unknown): “…that remain functional in individuals with motor impairments [29]. filter-bank common spatial pattern (FBCSP) was utilized as the primary analysis method due to its success in identifying discriminative frequency-specific features in motor imagery tasks [30, 31] and motor execution studies [32–34].”