基于惯性传感器的步数检测算法  

Step detection algorithm based on inertial sensor

作　　者：杜庆东[1] 范珏 DU Qingdong;FAN Jue(Software College,Shenyang Normal University,Shenyang 110034,China)

机构地区：[1]沈阳师范大学软件学院,沈阳110034

出　　处：《沈阳师范大学学报（自然科学版）》2021年第6期571-576,共6页Journal of Shenyang Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(31170380)。

摘　　要：行人轨迹推算(pedestrian dead reckoning,PDR)是一种无需参考基础设施的定位导航方法,由步数检测(step detection,SD)和步长估计(step length estimation,SLE)2个关键部分组成。在不同的步行模式中应用步数检测和步长估计是一个具有挑战性的问题。针对此问题,提出了一种基于使用智能手机的3种步行模式(即正常行走、原地踏步及快速行走)的鲁棒步数检测方法,通过基于动态时间扭曲(dynamic time warping)的结合峰值预测和过零检测方法来提高步数检测的精度。所提出的步数检测算法可以准确识别3种步行模式中每一步的起点和终点。针对不同的步行模式提出准确的步长估计模型,以提高步长估计的精度。实验结果表明,根据所提出的方法,步数检测的准确率约为97.8%,估计步行距离的误差约为4.0%,优于传统的步数检测和步长估计方法。As an infrastructure without reference to the positioning and navigation method,Pedestrian Dead Reckoning(PDR)is still a hot topic in the field of indoor positioning.Step detection(SD)and step length estimation(SLE)are two key PDR components.Applying SD and SLE to different walking patterns is a challenging problem.Focus on this problem,we propose a robust method SD(normal walking,standing and walking fast walk)based on the use of smart phone three walking mode.We propose a peak prediction and zero-crossing detection based on dynamic time warping to improve the accuracy of SD.SD presented accurately recognizes three pedestrian mode start and end of each step.An accurate SLE model for different walking patterns is made to achieve more accurate SLE.Experimental results show that,on average,SD accuracy rate is about 97.9%;the recognition accuracy is 98.4%;walking estimation error in distance is about 3.0%.It is better than the current method of SLE.

关 键 词：行人轨迹推算 步数检测 过零检测 峰值预测 

分 类 号：P315.69[天文地球—地震学]