Research Of Indoor Localization Based On Motion Mode Recognition And WIFI-PDR

In a closed building space environment,indoor positioning technology can quickly lock the indoor target position in real time,which is widely used in civil,commercial and even military applications.However,traditional indoor WIFI positioning methods rely on the collected Received Signal Strength Indicator(RSSI)fingerprint database,which is easily interfered by environmental factors,resulting in large positioning errors.While the Pedestrian Dead Reckoning(PDR)method,though with high short-term accuracy and free from environmental interference,will introduce cumulative errors as time passes.Meanwhile,some research shows that the positioning effect of PDR will be greatly affected by the different sensor data characteristics of intelligent devices in three holding postures(handheld,hand thrown,and under trouser pockets).To this end,this article combines indoor WIFI positioning and pedestrian track estimation methods to explore multiposture indoor positioning technology and its application in pedestrian handheld devices.The main research contents are as follows:(1)In order to reduce the positioning error introduced by the attenuation of WIFI signal and the fluctuation of intensity factor,an indoor ranging model based on RSSI is proposed.The GaussKalman filtering algorithm is used to filter out the abnormal data with small probability to reduce the environmental noise;a slope matching model is proposed which fits the curve in the short sliding window with a straight line to estimate the distance to solve the multipath interference problem.Experiments show that the ranging error of the proposed algorithm is lower than that of the logarithmic model and the Gauss-Kalman model.(2)Fuse the attitude angles calculated by the gyroscope and the magnetometer based on coordinate conversion and quaternion method to improve the accuracy of the PDR heading angle.K-Nearest Neighbor(KNN)classification is performed on the mean and variance of the single-step period under different postures.A KNN center distance algorithm is proposed to do the feature extraction and classification of the postures which effectively improves the posture recognition rate.After obtaining the posture data,deduce the corresponding step frequency information and calculate the real-time step length information based on the step length model of pedestrians with different characteristics such as genders and heights.(3)An improved particle filter algorithm MPF-RSSI is proposed,and it is dynamically weighted and fused with the positioning result of KNN center distance-PDR.Simulations and experiments show that the proposed algorithm has high robustness and universality,posture recognition rate,step recognition rate and positioning accuracy in the case of pedestrians with different step characteristics such as step length and step frequency,height and gender.When the positioning error is less than 2m,the algorithm in this thesis achieves a recognition rate of 90%,77%,74%and the maximum positioning root mean square error of 2.97m,3.47m,and 3.22m in the three postures of handheld,hand thrown,and under trouser pockets,respectively.Meanwhile,the positioning effects of the system in three scenarios of office buildings,shopping malls,and parking lots are compared,and the impact analysis of different influencing factors WIFI intensity,number of signal sources and operating distance on the positioning performance is given.