A second-generation non-invasive brain–computer interface (BCI) design for wheelchair control

A hands-free, wearable brain–computer interface (BCI) with fault tolerance is proposed for individuals with quadriplegia, enabling safe wheelchair control. It anticipates operational failures with fault-recovery methods to mitigate catastrophic outcomes. A multimodal system validates decoded signals for sensor consistency and confirms user intentions. The primary imaging system utilizes optical sensors to monitor brain signals, while a supplementary system employs accelerometers and gyroscopes to detect head movements. A third subsystem recognizes voice commands via built-in microphones. The primary control decodes the direction of movement, and the secondary confirms this through head tilting. A third redundancy uses voice commands to confirm or override wheelchair operations. Although cumbersome, head movements and voice commands are essential for users who are paralyzed and unable to press buttons in emergencies. Head motion sensors also detect headset slippage, halting operations immediately in the event of a headset drop. An adaptive voting system employing a majority-rule method filters out inconsistent outliers by utilizing historical patterns of consistency in the weight-sum voting process. To validate the interpretation of optical data collected from brain signals in the prefrontal cortex (PFC) using functional near-infrared spectroscopy (fNIRS), motor task experiments were conducted with human subjects performing horizontal hand movements in four orthogonal directions. The results indicated that oxy-hemoglobin (oxy-Hb) and deoxy-hemoglobin (deoxy-Hb) signals exhibited directional specificity, with responses reversing during opposing movements. It suggests that neural activity reflects the direction of movement. The consistency of the decoded results is verified by hemodynamic variables, with oxy-Hb and deoxy-Hb signals covarying inversely. Phase relationships between hemodynamic variables also changed depending on the direction of movement. An analysis revealed that the dynamics of vasodilation and vasoconstriction varied, indicating conditions where the oxygen supply could not meet metabolic demand in specific movement directions. The design allows users to choose a lightweight headband for cost-effective PFC monitoring or a helmet for whole-brain signal detection. The wearable design also provides wireless or wired headset options for communication with the controller located on the wheelchair. Future developments will address headset slippage challenges through adaptive signal processing, ensuring the commercial viability and reliability of the product.