In this paper, an active motion intention recognition technology in view of least squares support vector machine (LS-SVM) and zeroing neural network (ZNN) is proposed, and the continuous motion angles of knee joint and hip joint are successfully estimated from surface electromyography (sEMG). The processed sEMG signal is subjected to nonlinear mapping to gain the muscle activation which is applied as an input of the prediction model, and the output is the actual motion angle measured by the sensor. Compared with the standard SVM algorithm, the advantage of LS-SVM lies in the alteration of constraint conditions, which transforms the original quadratic programming (QP) problem into solving a set of linear equations. ZNN can be exploited to settle the converted equations and increase the calculation rate. In practice, it is difficult to avoid the interference of measurement noise on the recognition process, so a noise-suppressing zeroing neural network (NSZNN) is established and analyzed. Finally, the corresponding experiments show that the proposed method is able to accurately identify the joint angle, and when considering the noise in the solution process, it can eliminate the effect of noise on the estimation result to a certain extent. The numerical results reveal the method raised in this paper provides valuable reference for lower limb joint movement of patients.

本文提出了一种基于最小二乘支持向量机（LS-SVM）和归零神经网络（ZNN）的主动运动意图识别技术，并成功地利用表面肌电图估计了膝关节和髋关节的连续运动角度。 （肌电图）。对处理后的sEMG信号进行非线性映射以获得肌肉激活，作为预测模型的输入，输出为传感器测量的实际运动角度。与标准SVM算法相比，LS-SVM的优势在于改变了约束条件，将原来的二次规划（QP）问题转化为求解一组线性方程组。ZNN 可用于解决转换后的方程并提高计算速率。在实践中，由于测量噪声对识别过程的干扰难以避免，因此建立并分析了一种抑制噪声归零神经网络（NSZNN）。最后，相应的实验表明，所提方法能够准确识别关节角度，在求解过程中考虑噪声的情况下，可以在一定程度上消除噪声对估计结果的影响。数值结果表明本文提出的方法为患者下肢关节运动提供了有价值的参考。并且在求解过程中考虑噪声时，可以在一定程度上消除噪声对估计结果的影响。数值结果表明本文提出的方法为患者下肢关节运动提供了有价值的参考。并且在求解过程中考虑噪声时，可以在一定程度上消除噪声对估计结果的影响。数值结果表明本文提出的方法为患者下肢关节运动提供了有价值的参考。