功能电刺激腕关节运动的自适应控制技术研究

针对功能电刺激(FES)康复技术中肌肉的时变和高度非线性的特点,提出一种实时调整FES系统输出,实现康复患者自适应训练的方法。考虑到患者间和康复各阶段肌肉对电刺激响应能力的差异,利用神经网络良好的非线性逼近能力,采用神经网络辨识器辨识相关运动肌肉的FES模型,并在此基础上,利用神经网络控制器在线调节控制策略及刺激参数,提高FES系统的自适应能力。通过人体腕关节运动控制的FES实验,验证该方法具有较高的轨迹跟踪控制精度和较强的自适应能力。

Adaptive Neural Network Control of FES in Wrist Movements

The muscles are time-changeable and highly non-linear characteristic in functional electrical stimulation (FES) therapies so that it needs a self-adaptive control method to ensure the robustness of the system. Considering the different responding capable of each patient and each training phase for stimulating current, a self-learning control strategy was proposed in the paper. By a multiple level recurrent neural network identifies the real-time stimulating muscle states; the control algorithm could adjust the FES system output. Experimental results showed that the proposed method was robust and self-adaptive.