一种基于相关向量机的股四头肌收缩力量估计方法研究

针对可穿戴设备及共融机器人中的力/力矩测量需求，提出了一种基于相关向量机的人体股四头肌收缩力量估计方法，该方法具备采集设备安装方便、鲁棒性强且宜人性好等优点。通过采集人体股四头肌主要肌肉的MMG信号，提取平均绝对值MAV、平均功率频率MPF、样本熵SampEn及2个不同通道MMG信号之间的相关系数CC2Cs四个特征，利用基于稀疏贝叶斯理论的相关向量机算法RVM构建了MMG-肌肉收缩力量模型，并验证了所提方法的有效性和准确度。结果表明，同一参与者的模型估计结果的均方根误差RMSE为8.7%MVC（最大肌肉随意收缩力），决定系数R2为0.817，该方法是一种有效、适宜应用在可穿戴设备的人体股四头肌收缩力量估计方法。

An estimation method of quadriceps femoris contraction strength using Mechanomyography signal

In view of the measurement of force/torque for wearable devices and coexisting-cooperative-cognitive robots, a estimation method of human muscle contraction strength based on MMG siganl was proposed, which has the advantages of easy fixing, reliable usage and agreeableness. The MMG signal of the quadriceps femoris were collected, and then the mean absolute value (MAV), mean power frequency (MPF), sample entropy (SampEn), and correlation coefficients between two different MMG channels (CC2Cs) were extracted. On this basis, the model of MMG-muscle contraction strength was constructed using the relevant vector machine (RVM) based on Sparse Bayesian Modelling, the validity and accuracy of the proposed method was verified. The results indicated that the RMSE of the same-participant validation was 8.7% MVC and R2 was 0.817, which demonstrates it was an convenient, effective and suitable for wearable devices method for estimating human muscle contraction strength.