Title: A Real-Time BCI for Stroke Hand Rehabilitation Using Latent EEG Features from Healthy Subjects Show Chinese title

Title: 基于健康受试者潜在脑电特征的中风手部康复实时脑机接口

Abstract: This study presents a real-time, portable brain-computer interface (BCI) system designed to support hand rehabilitation for stroke patients. The system combines a low cost 3D-printed robotic exoskeleton with an embedded controller that converts brain signals into physical hand movements. EEG signals are recorded using a 14-channel Emotiv EPOC+ headset and processed through a supervised convolutional autoencoder (CAE) to extract meaningful latent features from single-trial data. The model is trained on publicly available EEG data from healthy individuals (WAY-EEG-GAL dataset), with electrode mapping adapted to match the Emotiv headset layout. Among several tested classifiers, Ada Boost achieved the highest accuracy (89.3%) and F1-score (0.89) in offline evaluations. The system was also tested in real time on five healthy subjects, achieving classification accuracies between 60% and 86%. The complete pipeline - EEG acquisition, signal processing, classification, and robotic control - is deployed on an NVIDIA Jetson Nano platform with a real-time graphical interface. These results demonstrate the system's potential as a low-cost, standalone solution for home-based neurorehabilitation. Show Chinese abstract

Abstract: 本研究提出了一种实时、便携的脑机接口（BCI）系统，旨在支持中风患者的手部康复。 该系统结合了一个低成本的3D打印机器人外骨骼和一个嵌入式控制器，可将脑信号转换为物理手部运动。 使用14通道的Emotiv EPOC+头戴设备记录EEG信号，并通过监督卷积自编码器（CAE）进行处理，以从单次试验数据中提取有意义的潜在特征。 该模型在公开可用的健康个体EEG数据（WAY-EEG-GAL数据集）上进行训练，电极映射适配以匹配Emotiv头戴设备布局。 在多种测试分类器中，Ada Boost在离线评估中达到了最高的准确率（89.3%）和F1分数（0.89）。 该系统还在五名健康受试者身上进行了实时测试，分类准确率在60%到86%之间。 完整的流程——EEG采集、信号处理、分类和机器人控制——部署在NVIDIA Jetson Nano平台上，并配有实时图形界面。 这些结果证明了该系统作为低成本、独立的家庭神经康复解决方案的潜力。