Scholarlink

Motor imagery electroencephalography (MI-EEG) decoding is challenged by noisy, non-stationary signals, high variability, and limited labeled data. We propose FA-GPNet, a unified framework that integrates Filter Bank Common Spatial Pattern (FBCSP) for multi-band spectral–spatial feature extraction, a deep autoencoder for nonlinear compression and noise reduction, and a Gaussian Process Classifier (GPC) for probabilistic, uncertainty-aware predictions. Unlike conventional FBCSP pipelines that rely on manual feature selection and deterministic classifiers, FA-GPNet replaces heuristic ranking with data-driven latent representation learning and leverages GP’s Bayesian framework for calibrated outputs. Under within-subject evaluation, FA-GPNet achieves 78.19% mean accuracy on BCI Competition IV-2b, surpassing strong traditional baselines and multiple deep networks, while remaining competitive with CapsNet. On the HCM-IU hand-binary MI dataset, FA-GPNet outperforms the well-optimized classical baseline. These results demonstrate that FA-GPNet offers a robust, reproducible, and efficient solution for MI-EEG decoding.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FA-GPNet: When Gaussian Process Meets Auto-Encoder and FBCSP - A Hybrid Model for Motor Imagery Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Trung M. Pham,
  • Hieu M. Pham,
  • Vi K. Nguyen,
  • Truong D. Tran,
  • Long S. T. Nguyen,
  • Duc Q. Nguyen,
  • Huong T. T. Ha,
  • Tho T. Quan

摘要

Motor imagery electroencephalography (MI-EEG) decoding is challenged by noisy, non-stationary signals, high variability, and limited labeled data. We propose FA-GPNet, a unified framework that integrates Filter Bank Common Spatial Pattern (FBCSP) for multi-band spectral–spatial feature extraction, a deep autoencoder for nonlinear compression and noise reduction, and a Gaussian Process Classifier (GPC) for probabilistic, uncertainty-aware predictions. Unlike conventional FBCSP pipelines that rely on manual feature selection and deterministic classifiers, FA-GPNet replaces heuristic ranking with data-driven latent representation learning and leverages GP’s Bayesian framework for calibrated outputs. Under within-subject evaluation, FA-GPNet achieves 78.19% mean accuracy on BCI Competition IV-2b, surpassing strong traditional baselines and multiple deep networks, while remaining competitive with CapsNet. On the HCM-IU hand-binary MI dataset, FA-GPNet outperforms the well-optimized classical baseline. These results demonstrate that FA-GPNet offers a robust, reproducible, and efficient solution for MI-EEG decoding.