【摘要】 为提高行人航迹推算（PDR）的定位精度，提出一种基于改进有限状态机（FSM）步态检测的PDR定位算法。通过引入基于K近邻的手机姿态识别算法，实现FSM步态检测算法中各项判断阈值的自适应调整，从而提高步态检测的精度和普适性，并针对不同手机姿态建立相应的Weinberg步长估计模型，增强PDR定位的鲁棒性。实验结果表明，改进步态检测算法不仅可以有效适应各种运动状态，而且各种手机姿态下的检测准确率均达到95%以上；相比基于波峰检测法的PDR模型，改进的PDR模型在四种手机姿态下的距离误差分别减小了91.4%、91.1%、84.6%和33.5%；而与基于传统FSM的PDR模型相比，距离误差则分别减小了61.5%、45.9%、78.5%和3.9%，有效提高了PDR定位精度。更多还原

【Abstract】 To improve the localization accuracy of pedestrian dead reckoning(PDR), a PDR location algorithm based on improved finite state machine(FSM) gait detection is proposed. By introducing a mobile phone pose recognition algorithm based on K-nearest neighbors(KNN), the adaptive adjustment of various judgment thresholds in FSM gait detection algorithm is realized, thereby improving the accuracy and universality of gait detection. Weinberg step size estimation models are established for different mobile phone poses, and the robustness of PDR localization is enhanced. The experimental results show that the improved gait detection algorithm not only effectively adapts to various motion states but also achieves a detection accuracy of over 95% for all phone postures. Compared to the PDR model based on peak detection, the improved PDR model reduces the distance error by 91.4%, 91.1%, 84.6% and 33.5% for four different phone postures, respectively.Compared to the traditional FSM-based PDR model, the distance error is reduced by 61.5%, 45.9%, 78.5% and 3.9%, effectively improving the PDR positioning accuracy.更多还原