SPD-DANN: An SPD manifold unsupervised domain adaptation method for cross subject motor imagery EEG decoding

Electroencephalogram (EEG) signals convey abundant physiological and psychological information. Decoding these signals is fundamental for brain-computer interfaces (BCIs) and medical rehabilitation. Nevertheless, the non-stationarity and inter-individual variability of EEG signals impede current models from achieving robust cross-subject generalization without expensive subject-specific recalibration, thereby restricting their practical deployment. Unsupervised domain adaptation (UDA) aims to improve generalization by minimizing distribution discrepancies between source and target domains. Recent studies treat individual subjects as distinct domains and leverage UDA to facilitate cross-subject EEG classification, learning domain-invariant features through discrepancy minimization or adversarial training. However, these conventional methods are primarily developed in Euclidean space, which is insufficient to capture the non-linear structural characteristics inherent in EEG data. To address this limitation, we propose a deep adversarial neural network on the Symmetric Positive Definite (SPD) matrix manifold, termed SPD-DANN, which facilitates the extraction of subject-invariant features via adversarial learning. Additionally, we design an SPD domain feature alignment loss and an SPD class prototype pair loss to simultaneously promote feature alignment across subjects and enhance discriminability within the feature space. Extensive experiments on four BCI datasets demonstrate that our method outperforms several state-of-the-art UDA techniques. Furthermore, the proposed loss functions are readily adaptable to broader unsupervised and semi-supervised domain adaptation frameworks.