基于表面肌电信号的人体步态事件快速识别方法

针对下肢康复训练机器人主动训练阶段患者运动检测实时性、准确性的需求，提出一种基于动态表面肌电信号的人体步态事件快速识别方法。首先，通过表面肌电信号产生过程数学建模及步态过程中肌肉活动规律分析，给出了基于表面肌电信号强度及其变化特征的步态事件感知原理；其次，以双腿股外侧肌动态表面肌电信号强度及其变化为特征，构建了用于识别支撑和摆动两个步态事件的自适应模糊神经网络模型。实验结果表明：该方法识别结果正确率达95.3%，对足跟触地和脚尖离地事件发生时刻进行识别的平均时间误差分别为21.4ms和24.5ms，同时证明，该方法对步态之间表面肌电信号的差异具有较强的鲁棒性。

Human Gait Events Fast Recognition Method via Surface Electromyography

Real-time and accuracy detection of human motion during active training stage were required for lower limb rehabilitation robot. A dynamic surface EMG based human gait events fast recognition method was proposed. Firstly，the surface EMG generation model was established and the skeletal muscle activity during gait was analyzed，the gait event perception method with surface EMG intensity and its variation was put forward. Then，an ANFIS model was built to recognize the supporting phase and swing phase, which used the dynamic surface EMG signals of vastus lateralis lie in the both of left and right thigh as the signal source.The experimental results show that the average correct rate may reach 95.3% compared with results detected from force plate，the average time errors for heel strike (HS) and toe off (TO) timing detection are 21.4ms and 24.5ms respectively. Moreover，the method proposed also shows a strong robustness against the surface EMG difference between gaits.