摘要
为了更快速精确地检测零速度,提出了一种自适应检测零速事件的行人导航方法。该方法首先根据惯性数据,以一个步态周期为一组,计算步行和跑步差异性特性标签;然后将计算出的标签通过支持向量机(SVM)判断运动状态;最后通过RNN神经网络使用原始惯性数据结合运动状态确定输出是否为零速度事件,降低了计算成本。通过不同运动状态的多个实验对该方法进行了评价,在135.6 m的步行,跑步和步行-跑步复合运动中,水平位置误差分别为0.748,0.593和1.054 m,闭合定位误差为0.551%,0.438%和0.777%。
In order to detect zero speed more quickly and accurately,a pedestrian navigation method based on adaptive detection of zero speed events is proposed.The method firstly calculates the different characteristic labels of walking and running based on the inertial data and one gait cycle as a group.Then the motion state of the calculated labels is judged by support vector machine(SVM).Finally,RNN neural network is used to determine whether the output is zero velocity event by combining the original inertial data with the motion state,which reduces the calculation cost.The method was evaluated by several experiments with different gaits.The horizontal position errors were 0.748,0.593 and 1.054 m,and the closed position errors were 0.551%,0.438%and 0.777%,respectively,in 135.6 m walking,running and walking-running combined exercises.
作者
刘思宇
崔建民
刘国栋
张溢阅
杨景宏
LIU Siyu;CUI Jianmin;LIU Guodong;ZHANG Yiyue;YANG Jinghong(Key Laboratory of information photon technology,Ministry of industry and information technology,School of optoelectronics,Beijing University of technology,Beijing 100081,China)
出处
《自动化与仪器仪表》
2024年第6期1-5,共5页
Automation & Instrumentation
基金
国家自然科学基金资助项目(62075012,61675025)。
关键词
行人惯性导航
零速修正
运动轨迹
运动分类
pedestrian inertial navigation
ZUPT
motion trajectory
motion classification