Integrated EEG acquisition and modeling system for neurofunctional biomarkers of covert motor intent

This study introduces a novel electroencephalogram (EEG) recording and modeling system that can be used to improve the inner speech decoding of Brain-Computer Interface (BCI) systems in neurorehabilitation and assistive technology. The framework is structured as an integrated sensingprocessing- modeling system that unifies acquisition, signal organization, neurofunctional interpretation, and model-level computation into a coherent end-to-end pipeline. The data was gathered with the help of motor imagery tasks, namely, left hand, right hand, and leg movement intentions, with the help of a carefully controlled visual-cue paradigm using the Emotiv EPOC X headset. Neu- rofunctional analysis showed that there was a consistent fronto- motor coupling, with the AF3-FC5 (left prefrontal cortexleft motor cortex) channel pair having an R 2 value of up to 66.6%, compared to posterior pairs like AF3-O1 (left prefrontal cortex - left occipital lobe), which had a maximum of 0.2%. Mapping of channel-wise activation showed that AF3, FC5, and. F4 (right dorsolateral prefrontal cortex) were dominant in 90 to 100% of participants. Greater O2 (right occipital lobe) and lower FC6 (right motor cortex) activity indicate rightbiased visual and motor activity. The study also introduces Novel Gated Recurrent Convolution Neural Network (GRCNet), a GRU (Gated Recurrent Unit)-augmented convolutional architecture, facilitating unified spatiotemporal learning without the need for deep recurrent stacks. Experimental results indicate test accuracies stabilizing around 65–70%, outperforming the baseline architectures that were evaluated in this study, which suggests potential representational advantages that warrant further investigation on larger datasets. This work offers an initial step toward integrated and interpretable EEG-based BCI pipelines, with scope for further validation on larger and more diverse cohorts.