具有自然柔性与本体感性能力的上肢肌电假肢技术研究

Most of existing myoelectric prostheses are one-way discretely controlled to track the user’s motion types identified via multi-channel surface electromyography (sEMG). The interaction barrier is generated due to that the prostheses are unable to simulate flexible movements of human limbs and the users also can’t perceive prosthesis movement. The neural-interaction interface technology would help to realize human-prosthesis integration interaction. However, the invasive nerve electrode should not be used widely due to the harm to human body. This project investigates the neuromechanism of motion production, transmission and feedback, and then study how to realize the two-way neural interaction technology between human and prostheses via an non-invasive way. The motor nerve potential decoding method based on sEMG decomposition is proposed and a model is built to recognize human motion and to estimate the stiffness of human joints. In order to simulate the intrinsic flexibility of human joints, the flexible joint of prosthesis with variable stiffness is designed. Then, the joint stiffness is actively controlled by using the estimation of human movement intention. Meanwhile, the project analyzes the feedback perception mechanism with respect to human-limb movements, and studies the neural coding technology of the feedback perception via non-intrusive external physical stimulation. Afterwards, a feedback perception model of prosthetic movements is built to realize two-way neural interaction between human and prosthesis. With the proposed methods, myoelectricity prosthesis with natural flexibility and self-perception can be developed to reproduce or to enhance the movement ability of imperfect limbs.

现有肌电假肢一般利用多通道表面肌电信号(sEMG)识别使用者动作类型，实现对假肢的单向离散动作控制，无法模拟肢体运动自然柔性的特点，使用者也无法感知假肢运动状态，极易产生人机交互隔阂。神经交互接口技术有助于实现人与假肢的共融交互，但是侵入式神经电极伤害人体，不适合推广应用。本项目探索运动产生、传输及反馈的神经机制，研究通过非侵入方式实现人与假肢双向神经交互技术。提出基于sEMG分解的运动神经电位解码方法，建立识别动作与估计人体关节刚度的模型；设计假肢可变刚度柔性关节，基于估计的人体运动意图，实现对柔性关节刚度主动控制，模拟人体关节内在自然柔性；同时，分析人体对运动反馈的感知机理，研究通过非侵入外部物理刺激模拟反馈感知的神经编码技术，建立人体对假肢运动反馈感知模型，从而实现人与假肢的双向神经交互。利用本课题提出方法有利于开发具有自然柔性和本体感应能力的肌电假肢，复现或增强残缺肢体的运动功能。

为了提高肌电控制假肢的稳定性和使用便捷性，进一步提升假肢的自然柔性和本体感应能力，本项目深入探索肌电产生及传输机理，提取直接反应运动本质的神经电位信号，即活跃动作电位序列(MUAPT)，利用MUAPT识别人体运动意图；设计模块化柔性关节，并研究基于人体意图控制的关节刚度在线调整策略；进一步研究不同外部物理刺激的神经反馈感知机理，建立人体实时感知假肢运动状态及交互环境信息的模型。将上述方法/技术集成，构建上肢肌电假肢实验平台，通过人机共融试验验证相关方法/技术。.研究主要从控制假肢手的人体意图识别技术研究、柔顺假肢手本体核心技术研究、假肢手控制鲁棒性提升方法研究、假肢手反馈及其集成研究4个方面展开。详细包含：.1）.基于sEMG的人体意图识别方法研究：基于机器学习的方式，利用sEMG识别人体控制假手手的意图；.2）.智能假肢柔性驱动与柔性关节研究：主要包括基于形状记忆合金的假肢驱动动态建模与补偿控制研究和基于阿基米德螺旋线型平面涡卷弹簧的假肢柔性关节研究两方面；.3）.基于MUAPT与APA方法的假肢控制鲁棒性增强方法：主要包括基于小波变换的MUAPT提取方法研究和基于极坐标矫正电极偏移下的鲁棒性提升方法研究；.4）.引入反馈下的假肢系统集成与控制方法研究：主要包括基于“竖形条”的视觉感知反馈下假肢手控制研究和假肢手控制系统的软硬件集成研究。