Image Segmentation-P300 Selector: A Brain–Computer Interface System for Target Selection

Brain–computer interface (BCI) systems, such as the P300 speller, enable patients to express intentions without necessitating extensive training. However, the complexity of operational instructions and the slow pace of character spelling pose challenges for some patients. In this paper, an image segmentation P300 selector based on YOLOv7-mask and DeepSORT is proposed. The proposed system utilizes a camera to capture real-world objects for classification and tracking. By applying predefined stimulation rules and object-specific masks, the proposed system triggers stimuli associated with the objects displayed on the screen, inducing the generation of P300 signals in the patient’s brain. Its video processing mechanism enables the system to identify the target the patient is focusing on even if the object is partially obscured, overlapped, moving, or changing in number. The system alters the target’s color display, thereby conveying the patient’s intentions to caregivers. The data analysis revealed that the self-recognition accuracy of the subjects using this method was between 92% and 100%, and the cross-subject P300 recognition precision was 81.9%–92.1%. This means that simple instructions such as “Do not worry, just focus on what you desire” effectively discerned the patient’s intentions using the Image Segmentation-P300 selector. This approach provides cost-effective support and allows patients with communication difficulties to easily express their needs.