混合脑电信号及视觉信息的智能轮椅人机交互系统

针对单一脑电信号人机交互系统中受试者长时间运动想象过程中精神极易产生疲乏，导致脑电信号有用信息量不足造成系统误识别的问题，本文提出一种视觉信息辅助脑电信号的人机交互系统。该系统通过在脑电信号实时操作中不断地识别眼睛状态，产生一种新的样本更新策略，更新的视觉信息作为系统的反馈，对人机交互闭环控制系统起到有效的校正作用。通过在智能轮椅平台走一个“8”字形固定轨迹的实验，实验结果表明：视觉信息的加入有效地避免了单一脑电信号控制智能轮椅由于疲乏问题造成的误识别问题，具有较好的稳定性和鲁棒性，表明了该人机交互方法的可行性。
To address error recognition problems created by mental fatigue when a human subject partakes in a single motor imager process for a long period of time, a visual information assisted EEG (electroencephalograph) human-machine interactive control system was proposed. The system produces a new sample updated strategy, with the‘state’of the eyes being recognized by the improved Adaboost algorithm in real-time and the recognition result being used to decide which EEG signal to update as the model parameter for human-machine interaction. An experiment on controlling an intelligent wheelchair off a fixed trajectory with a‘8’glyph was undertaken. The results show that visual information is adopted effectively by the intelligent wheelchair users to avoid the fatique-related error recognition problem with good levels of efficiency; thus proving that the interactive method is feasible