卒中造成的神经损伤会引起患者感觉障碍，进而阻断康复阶段的运动再学习。以人工神经反馈的手段弥补感觉障碍已证明可改善患者的运动功能。但由于尚未建立神经反馈下对运动学习的持续观测，使康复方案无法根据学习进程灵活调整。我们前期建立了一套基于肌电的实时神经反馈技术能较好改善患者即期运动功能。其下一步关键科学问题是根据卒中患者运动学习的速率建立自适应动态反馈方案，实现卒中个性化智能康复。本课题将：1）研究能否增强卒中患者运动学习中的肌肉感知，开发以随机微分方程模型将肌电实时转为体表震动的技术；2）确定神经反馈的肌肉靶点，捕捉受损肌肉与正常肌肉的噪声水平差异；3）研究神经反馈下卒中患者的运动学习，基于运动学习速率调整神经反馈增益。神经反馈下的运动学习将以人体运动学中“速度—精确度置换率”等指标进行客观评定。本课题的结果将为卒中康复训练提供新途径，可迅速转化至康复临床，并对卒中后运动学习理论提供新证据。

Post-stroke neural impairments may cause sensory deficit in survivors, and the sensory deficit could potentially block motor learning during rehabilitation. It is suggested that motor functions can be improved by compensating sensory deficit using neurofeedback, a technology of sensory augmentation. The motor learning under neurofeedback, however, has not been continuously monitored, therefore it is difficult to adjust therapeutic plans commensurate to learning processes. In previous work, we developed the technique of real-time neurofeedback using electromyography to improve the impaired motor functions after short-term training. The next key question is how to create an adaptive control policy compatible with the rate of motor learning in stroke survivors, which will provide individualized and intelligent rehabilitation. In this project, we will: 1) study whether the awareness of muscle activation can be augmented in stroke survivors, which requires converting surface electromyography using stochastic differential equation to mechanical vibration on human body; 2) characterize the difference in noise level between normal and abnormal muscles, so as to locate target muscles for neurofeedback; 3) study the adaptive control policy and physiological mechanism of neurofeedback by monitoring the motor learning process. Motor learning under neurofeedback will be evaluated using kinesiology benchmarks including “speed-accuracy tradeoff”. This study will provide new approaches for stroke rehabilitation, technologies that are easy to translate into clinical practice, and new data for theories of post-stroke motor learning.