摘要
根据不同路况条件和典型步速的笛卡尔积组合,利用装配在残肢侧的陀螺仪、加速度计和足底前后的压力传感器的信息,通过相关性系数分析、传感器融合、隐马尔可夫模型的方法,判断假肢使用者的运动意图.以健肢运动状态为参考值,利用迭代学习控制分别建立不同路况和步速情况下的控制知识数据库.通过传感器的关键状态变化信号驱动有限状态机状态转换,输出控制知识库中的控制量,实现假肢膝关节在不同路况、步速条件下对步态相位的控制.针对控制过程中出现的输出量实时偏差,采取了在线校正措施.对于有限状态机输出控制数据序列在时间同步上的超前和滞后问题,采取了相应的保持和补偿措施.结果表明,经隐马尔可夫模型处理后路况判断准确率可提升到91.7%,基于数据驱动的无模型控制方法能够实现对不同路况、步速下假肢步态的有效控制.
According to the combination of different terrains and walking speeds in the way of Cartesian product,a motion intention recognizer for amputee was presented.The sensor system was composed of an accelerometer,a gyroscope mounted on the prosthetic socket,and two pressure sensors mounted under the sole.The motion intention was inferred by intra-class correlation coefficient,sensor fusion and hidden Markov model.And a flexible iterative learning control(ILC) was proposed to build an experience database for the control of knee joint in prosthesis.And the motion state of the healthy knee was set as the learning sample in ILC.Furthermore,the sensor signals of the state transition were used to drive a finite state machine(FSM).The control experience in the knowledge database was output to control the stride phase according to the terrain,and speed.Besides,an online correction was adopted to reduce the real-time errors in the output axis.Moreover,to regulate the output sequence lead and lag in time axis,an output holder and a compensator were used.The experimental results show that the accuracy of the terrain recognition using the hidden Markor model is improved by 91.7%.Thus,the model-free control method is effective for prosthesis gait control of prosthesis according to the terrain and speed.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2017年第6期1107-1116,共10页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(61703135
61773151)
河北省自然科学基金青年基金资助项目(F2016202327)
关键词
假肢
运动意图
迭代学习控制
有限状态机
prosthesis
motion intention
iterative learning control
finite state machine
