基于目标导向MI-EEGas的上肢运动康复方法双向适应性研究

MI-BCI康复方法是基于运动想像脑电信号(MI-EEG)和脑机接口技术(BCI)的主动式康复方法，在肢体运动功能康复中具有广阔应用前景，而脑-机间的双向适应性是影响康复效果和制约临床应用的关键。课题针对脑卒中后单侧上肢运动康复问题，从交叉学科的角度，拟以健康动机和Skinner强化学习理论为指导，设计目标导向MI-EEGas个性化采集实验，以充分调动患者内在动力，激励其不断强化学习并及时修正和促进健康行为，提高脑对机的适应性；以流行学习方法和基于正交本征模态分解的希尔伯特-黄变换为主要工具，确保多域特征信息的完整性与自适应性，并基于细胞不对称分裂原理和Adaboost技术研究自组织特征映射网络(SOM)动态结构自调整策略和集成SOM自学习分类器设计方法，增强机对脑的适应性。预期研制一种便携式上肢在线康复原型系统，实现患肢运动想像与实际训练实时交互，达到改善康复效果和促进临床应用的目的。

MI-BCI is an active rehabilitation method with motor imagery EEG (MI-EEG) and brain computer interface (BCI) technology, and has a wide application prospect in limb motor function recovery. In MI-BCI,the dual adaptivity between brain and computer is the key issue which impacts the recovery effect and limits the clinical application. In the project, only unilateral upper limb dyskinesia after sroke is considered and some research will be done for its function rehabilitation problem based on interdisciplinary science. Under the guidance of healthy motivation and Skinner reinforcement learning theory, the goal-directed and personalized collection experiment of MI-EEGas, which is derived from the affected side, will be designed to arouse the patient's intrinstic motivation sufficiently and encourage him (or her) to correct and promote his (or her) healthy action in time by reinforcement learning. So the "brain vs computer" adaptivity will be improved. Then, the manifold learning and Hilbert-Huang Transform based on orthogonal intrinsic mode function decomposition will be used to ensure the integrity and adaptability of feature information. Furthermore, according to the principle of asymmetric cell division and Adaboost technology, the dynamic structure self-adjustment strategy of self-organizing feature map neural network (SOM) will be researched, and then the self-learning SOM ensemble classifier will be designed. Therefore the "computer vs brain" adaptability will be enhanced. At last, a portable on-line original upper limb rehabilitation system will be obtained, and the motor imagery and real training from the affected upper limb will be interacted. This is helpful for improving the motor function recovery effect and promoting the clinical application.