Frequency–Spatial Cognitive Intelligence: A Deep Learning Framework for Advanced BCI Systems

Abstract

Brain-computer interfaces (BCIs) allow direct communication between the brain and external devices by decoding neural signals, primarily from the EEG. Motor imagery-based BCIs face challenges due to non-stationary, noisy, and spatially complex signals. This paper introduces the Spatially Enriched Cognitive Intelligence (SECI) framework, which leverages deep neural networks to capture spatial and temporal dependencies in EEG signals while incorporating frequency-aware attention mechanisms. SECI integrates a spatial encoder, cognitive feature extractor, and temporal reasoning module to improve classification accuracy for motor imagery tasks. Evaluations on BCI Competition IV-2a and PhysioNet datasets demonstrate that SECI achieves an average classification accuracy of 89.8%, an F1-score of 0.88, and a Cohen’s Kappa of 0.86, outperforming conventional CNN and hybrid CNN-LSTM models by 5–8%. These results establish SECI as a robust, interpretable, and high-performing framework for assistive and real-time BCI applications.