Brain signal decoding is essential for unraveling the intricate neural processes in the human brain. Various monitoring modalities, including electroencephalography (EEG), mag-netoencephalography (MEG), and electrocorticography (ECoG), provide valuable insights into cortical activities. However, these modalities exhibit variations in waveform, frequency resolution, and signal properties, posing challenges for unified decoding methods. This paper introduces a compact end-to-end multi-branch convolutional neural network (CNN) architecture designed to decode brain signals from diverse modalities. The model integrates designated feature extractors, ensuring versatile performance across modalities and paradigms. The architecture addresses data variability, expands receptive fields, and enhances generalizability, surpassing existing brain signal decoders on two EEG datasets (achieving accuracies of 77.89% and 72.25%), one MEG dataset (52.97%), and one ECoG dataset (93.31%). The proposed model exhibits robustness without requiring manual tailoring and signifies a substantial advancement in the general decoding of brain signals.

Recognition of electroencephalographic (EEG) signals highly affect the efficiency of non-invasive brain-computer interfaces (BCIs). While recent advances of deep-learning (DL)-based EEG decoders offer improved performances, the development of geometric learning (GL) has attracted much attention for offering exceptional robustness in decoding noisy EEG data. However, there is a lack of studies on the merged use of deep neural networks (DNNs) and geometric learning for EEG decoding. We herein propose a manifold attention network (mAtt), a novel geometric deep learning (GDL)-based model, featuring a manifold attention mechanism that characterizes spatiotemporal representations of EEG data fully on a Riemannian symmetric positive definite (SPD). The evaluation of the proposed mAtt on both time-synchronous and -asyncronous EEG datasets suggests its superiority over other leading DL methods for general EEG decoding. Furthermore, analysis of model interpretation reveals the capability of mAtt in capturing informative EEG features and handling the non-stationarity of brain dynamics.