Scholarlink

In this study, we proposed a deep learning model combining convolutional neural networks (CNN) and long short-term memory networks (LSTM) for classifying four categories of motor imagery EEG signals (left hand, right hand, foot, and tongue). The CNN layer extracts extracting spatial features between EEG signal channels, and the stacked LSTM layer is used to capture the temporal dependency of the signal. By combining spatial and temporal modeling, the CNN-LSTM structure can effectively characterize the spatial distribution and dynamic patterns of EEG signals. Specifically, compared with single-structure ANN or CNN or LSTM models, the proposed method achieving a validation accuracy about 80.6% and a stable loss around 6.2%. It shows higher classification accuracy and stability, which is highly suitable for motor imagery tasks based on EEG signals, and exhibits favorable lightweight properties.

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

Robust Four-Class Motor Imagery EEG Classification Using Lightweight CNN-LSTM

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Haoran Zheng

摘要

In this study, we proposed a deep learning model combining convolutional neural networks (CNN) and long short-term memory networks (LSTM) for classifying four categories of motor imagery EEG signals (left hand, right hand, foot, and tongue). The CNN layer extracts extracting spatial features between EEG signal channels, and the stacked LSTM layer is used to capture the temporal dependency of the signal. By combining spatial and temporal modeling, the CNN-LSTM structure can effectively characterize the spatial distribution and dynamic patterns of EEG signals. Specifically, compared with single-structure ANN or CNN or LSTM models, the proposed method achieving a validation accuracy about 80.6% and a stable loss around 6.2%. It shows higher classification accuracy and stability, which is highly suitable for motor imagery tasks based on EEG signals, and exhibits favorable lightweight properties.