基于MPSO-BP神经网络方法的人体步态识别

Human gait recognition based on MPSO-BP neural network method

为提高人体下肢步态相位识别准确率以实现外骨骼机器人控制,采用一种改进的粒子群优化MPSO-BP神经网络方法识别不同运动模式下的人体步态相位。通过自适应调整学习因子构造MPSO-BP神经网络分类器,以多种传感信息组成的特征向量样本集训练神经网络分类器,用于识别人体下肢在平地行走、上楼梯和起坐三种典型运动模式下的步态相位。实验结果表明,MPSO-BP神经网络分类器能有效识别三种不同运动模式的步态相位,识别准确率均达到96%以上,识别性能优于传统的BP神经网络模型和粒子群优化神经网络模型。

To improve the accuracy rate of human gait phase recognition for controlling the exoskeleton robot, an approach based on Modified Particle Swarm Optimization algorithm-Back Propagation（MPSO-BP）neural network is utilized to divide three types of gait into different phases. Firstly, the MPSO-BP neural network classifier is constructed through regulating the learning factor adaptively, and then the classifier is trained using sample set containing multisensor information. Secondly, test the classifier on gait phase recognition in three types of human gait including walk,upstairs and sit-down. The experimental results show that the MPSO-BP neural network classifier can successfully increase the accuracy rate up to averaged 96% above, which is superior to the BP neural network and the particle swarm optimization BP neural network methods.