A simple deep transfer learning model with feature alignment block for motor imagery decoding

To address data scarcity and distribution shifts in motor imagery electroencephalogram (MI-EEG) based brain computer interface, we propose a 1-dimensional convolution-based deep transfer learning model with embedded Feature Alignment block (1DC-DTL-FA) in this article. It integrates multi-stage feature extraction, classification, and FA block. Unlike complex models, it utilizes Neural Architecture Search (NAS) to automatically locate the optimal FA position in Euclidean space Evaluated on BCI 2000 and BCI IV2a datasets, 1DC-DTL-FA achieved superior accuracies of 89.80% and 82.96%. The results demonstrate that this simple architecture effectively handles complex feature extraction and online alignment, outperforming state-of-the-art models in MI-EEG decoding.