基于GWO-SVM的下肢假肢穿戴者骑行相位识别

针对下肢假肢穿戴者骑行相位识别的问题，提出基于灰狼算法优化的支持向量机(GWO-SVM)分类模型. 建立下肢多源信息系统，采集膝关节、踝关节的加速度信号以及膝关节角度信号. 应用奇异值分解，对采集到的信号进行降噪处理. 在对信号进行降噪处理之后，为了避免单一信号不确定的影响，从数据冗余角度，选取各信号的特征点，开展归一化处理，组成多维特征向量，作为SVM分类模型的输入. 为了能够进一步提高分类精度，加强全局优化能力，利用GWO算法对核参数进行优化. 通过与PSO-SVM分类模型、GA-SVM分类模型对比表明，基于GWO优化的SVM分类模型对骑行相位的识别率为94%，高于其他方法优化的SVM分类模型.

Bicycle riding phase recognition of lower limb amputees based on GWO-SVM

An approach based on gray wolf algorithm optimization and support vector machine was proposed aiming at the problem of identifying the riding phases of lower limb prosthetic wearers. A multi-sensor system was constructed for the motion data collection. Then the acceleration signals of the knee joint, ankle joint and the angle signals of the knee joint on the prosthetic side were collected. Singular value noise reduction was used to reduce the noise of the collected signal. Then feature points of each signal were selected and normalized from the perspective of data redundancy. These motion feature point signals formed a multi-dimensional feature vector as the input of SVM classification model, which solved the problem of the uncertain influence of a single signal. The gray wolf algorithm optimized support vector machine kernel parameters, which not only improved the classification accuracy of the recognition, but also enhanced the global optimization ability. The support vector machine model optimized by the gray wolf algorithm has an accuracy rate of 94% for bicycle riding phase recognition, which is higher than the support vector machine model based on particle swarm optimization and the support vector machine model optimized based on genetic optimization algorithm.