A multiscale siamese convolutional neural network with cross-channel fusion for motor imagery decoding

The proposed network achieves an average accuracy of 87.36% and 87.33% on public BCI Competition IV 2a and 2b datasets.

• A multiscale Siamese convolutional neural network with cross-channel fusion is introduced for motor imagery electroencephalography classification.
• The network features three components: Siamese cross-channel fusion streams, a similarity module, and a classification module.
• Cross-channel fusion modules within the network enhance the representation of multiscale temporal features.
• A joint training strategy combining the similarity and classification modules aims to optimize network performance.
• The proposed method outperforms existing state-of-the-art techniques in classifying MI-EEG signals.

AI simplified