Incorporation of a coding scheme based on neuronal firing patterns and distributions improves noninvasive brain control of robotic devices

Brain–computer interfaces (BCIs) represent a breakthrough that enables individuals to restore motor function and extend their ability to interact with external environments. Although invasive BCIs have demonstrated remarkable performance in neuroprosthetic control by directly decoding intracranial neuronal activities related to motor intention, their application is inevitably limited by surgical risks. Meanwhile, although noninvasive BCIs based on scalp electroencephalography (EEG) signals are theoretically safer for broader use, they are limited by inefficient and poor movement control performance, mainly owing to the lack of robust and efficient mapping between fine movement intention and EEG signals. To address this gap, we propose emulating the natural neuronal coding mechanisms of the brain to encode movement intention into EEG signals. We found that EEG potentials modulated by visual directions exhibit both single-channel cosine-like tuning and multichannel population coding patterns, resembling the motor directional tuning distribution of neuronal firing. Building on these findings, we encoded intended movement directions into EEG potentials modulated by visual directions and introduced a high-order multiscale discriminative analysis algorithm to decode the intended directions. We evaluated our approach through rigorous experiments involving cursor control, ground vehicle navigation, and quadcopter operation. Results from twenty-six participants demonstrated that our system could reach a median squared tracking correlation of 0.53, outperforming the best contemporary noninvasive BCI by 400% and even rivalling the performance of invasive counterparts. Notably, owing to its excellent performance, our approach could be applied in a high-cognitive-demand task, i.e., controlling a quadcopter pursuing, targeting, and photographing a moving vehicle. Thus, we established a neuronal direction-based coding framework to enhance EEG-based movement control, which represents the first application of neuronal coding principles to noninvasive BCIs. This approach narrows the performance gap with invasive systems, providing a safe and practical alternative to high-performance neuroprosthetics and enabling accurate and timely movement control in both daily-life interaction systems and unmanned operation scenarios.