Scholarlink

Electroencephalogram (EEG) signals are widely used in brain-computer interface (BCI) systems for their high temporal resolution and non-invasiveness. However, their non-stationarity, low signal-to-noise ratio (SNR), and complex spatiotemporal patterns pose significant challenges for classification. To address this, we propose a deep adjustable convolutional neural network that integrates the Squeeze-and-Excitation (SE) attention mechanism. The network features flexible depth adjustment, making it well-suited for complex tasks, particularly SSMVEP signal processing. Experiments on a self-built SSMVEP dataset and the public 12JFPM_SSVEP dataset demonstrate that our method surpasses mainstream models in accuracy, robustness, and scalability, showing strong potential for multi-class EEG classification and advancing BCI development.

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

Adaptive Network Design for SSVEP/SSMVEP Classification via SE and Configurable Convolutions

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yichen Lin,
  • Xiuyuan Wu,
  • Xinyang Du,
  • Haoran Zhang,
  • Wenke Lu,
  • Yu Zhu,
  • Zengle Ren,
  • Pengjie Qin,
  • Jinke Li,
  • Yue Ma

摘要

Electroencephalogram (EEG) signals are widely used in brain-computer interface (BCI) systems for their high temporal resolution and non-invasiveness. However, their non-stationarity, low signal-to-noise ratio (SNR), and complex spatiotemporal patterns pose significant challenges for classification. To address this, we propose a deep adjustable convolutional neural network that integrates the Squeeze-and-Excitation (SE) attention mechanism. The network features flexible depth adjustment, making it well-suited for complex tasks, particularly SSMVEP signal processing. Experiments on a self-built SSMVEP dataset and the public 12JFPM_SSVEP dataset demonstrate that our method surpasses mainstream models in accuracy, robustness, and scalability, showing strong potential for multi-class EEG classification and advancing BCI development.