Research On Indoor Localization Fusion Of Mobile Phone Sensors And Location Fingerprint

Social and commercial interests in location-based services are significantly increasing due to the rise in smart devices and technologies.The GNSS have long been employed for locationbased services to navigate and determine accurate and reliable location information in outdoor environments.However,GNSS signals cannot penetrate buildings and complex indoor environments,GNSS cannot provide reliable location-based services in indoor environments.The deployment of MEMS sensors in smartphones have initiated new opportunities and challenges for the industry and academia alike.The demand and potential of location-based services in indoor environments force researchers to study more reliable,accurate,and low-cost indoor localization methods.Pedestrian Dead Reckoning(PDR)method and location fingerprinting method are commonly used in smartphone-based indoor localization technologies.PDR can perform real-time and continuous positioning of users.However,The use of PDR for positioning will cause large directional drift and cause serious cumulative errors.The location fingerprint method cannot perform continuous precise location,and the poor distinguishability of the location fingerprint often leads to incorrect location estimation.This paper studies the fusion of mobile phone sensors and location fingerprints to provide users with a high-precision,continuous indoor localization method.The research aims to suppress the cumulative error of PDR,raise the localization accuracy of the position fingerprint method,and improve the distinguishability of the fingerprint.The main research contents are as follows:1.A step-counting algorithm based on adaptive time window.For PDR algorithms,accurate step counting has always been a core problem to be solved,and accurate step counting is also a prerequisite for using PDR algorithms.The existing step counting algorithms cannot effectively solve the problem of the diversity of user walking patterns and mobile phone holding styles.Therefore,an adaptive time window step counting algorithm based on peak detection is proposed.The algorithm counted the steps by detecting and verifying,used double filtering to preprocess the original acceleration,and designed an adaptive time window based on the peak and valley time difference to eliminate false peaks,used the variance and standard deviation to verify the peak,finally.The experimental results show that the algorithm has a better improvement in the average step count accuracy under different walking patterns and different mobile phone holding styles,which is better than the commonly used methods based on peak detection and the current popular commercial step calculation applications,and its accuracy and adaptability Significantly improved.2.A fusion localization algorithm of particle filter and PDR.PDR faces two challenges.The first is direction drift,which leads to the phenomenon that the estimated walking trajectory passes through the wall,causing serious cumulative errors.Second is the step length estimation.There are multiple states involved in the walking process,and a single model cannot cover all walking information.In this paper,particle filtering is used to correct the direction in the walking process,and the walking direction is corrected by setting the reachable area of the particles,the direction attribute of the particles and the angular deviation.The Light GBM model is used to estimate the walking step length,and an accurate step length estimation model can be obtained by using data of different states for training.The experimental results show that the PDR using particle filter for direction correction effectively solves the problem of direction drift and reduces the accumulated error.In addition,the Light GBM model can accurately estimate the step length during walking in real time.3.A fusion localization algorithm of magnetic field fingerprint and PDR.Aiming at the problem of particle extinction and adaptability.This paper proposes an indoor localization algorithm that combines magnetic field fingerprints and PDR.The fusion localization algorithm has two steps.First,the direction of PDR is corrected by particle filtering,and the particle weight in the particle filter is updated with the strength of the magnetic field.Second,the acquired magnetic field strength is processed to obtain new features.The magnetic fingerprint database established by the new features can improve the low distinguishability of magnetic fingerprints,and then use the K-nearest neighbor method to estimate the location.Finally,the Extended Kalman Filter(EKF)fused the PDR positioning result after direction correction and the positioning result of the magnetic field fingerprint method to obtain the final position estimate.Experimental results show that the mean localization error is 0.83 meters,and the RMSE is 0.90 meters.It can provide more accurate and continuous real-time position estimation,and the proposed algorithm improves the distinguishability of magnetic field fingerprints.Adaptability and robustness.The algorithm has good adaptability and robustness.