| Location-based services have become the infrastructure of life and various applications.The location-based services provided by the Global Positioning System(GPS)have shown extremely high availability outdoors with good signal strength,bringing users It is very convenient,but in an indoor closed environment,due to the complex environment and the reflection and occlusion of satellite signals by walls,the GPS system cannot provide users with a real-time and accurate indoor positioning solution.The position fingerprint positioning method and the sensor-based positioning method Fusion for localization is a promising method to solve such problems.This thesis mainly studies the Wi Fi location fingerprint positioning method and the indoor positioning method based on inertial sensor.Through the two-step initial calibration of the sensor,the classification of different activity states,the expansion of the number of samples and other ways,the problems of low accuracy,high delay and vulnerable to noise of the positioning system are solved.Aiming at the problem of inaccurate signal acquisition caused by the RSSI signal being easily affected by multipath effect and random noise in the offline construction of Wi-Fi fingerprint database,the RSSI signal feature extraction algorithm based on the Gaussian filter model is used to filter the noise in the signal.Get more reliable fingerprint data.Aiming at the problem of high positioning delay caused by the large amount of fingerprint database data,the offline AP selection algorithm is used to select APs with obvious characteristics,which reduces the size of the fingerprint database and reduces the average positioning delay by 27.3%.In order to reduce the construction cost of fingerprint database,a sparse sample expansion method based on Gaussian process regression is adopted to expand the RSSI samples with obvious eigenvalues.Under the condition of constant workload,the number of samples is greatly increased.In view of the inaccurate output results of inertial sensors in the initial state,an initial two-step calibration method is established.The information is obtained by using the sensors on the smart phone,and then the error information of each sensor is processed by using the filtering algorithm to obtain the deviation from the real coordinate system,so as to provide guarantee for subsequent fusion positioning.Aiming at the problem that users have great differences in step frequency information and step size information under different activity states,the step detection algorithm based on pedestrian activity analysis is adopted,and the step frequency and step size estimates are adjusted according to different activity categories.The average recognition accuracy is 98%,and the recognition accuracy in complex activity state detection is 92.35%,which are better than the step detection algorithm used in Ipin competition system.This thesis constructs an experiment in a typical office scene.The results show that the average positioning accuracy can reach 1.21m,and 73%of the time positioning errors are less than 1m,which effectively improves the indoor positioning accuracy.In addition,an indoor positioning system based on Wi-Fi fingerprint and inertial sensors is designed,and the established system is integrated to form a complete indoor positioning model,which can obtain indoor position information with high efficiency,high precision,low cost and low delay. |
PDF Full Text Request