Research On Indoor Positioning Algorithm Of Smartphone Based On Inertial Sensor

Benefit from the growing of semiconductor technology,micro-sensors such as acceleration sensors,gyroscopes,and magnetometers have been widely integrated in smartphones,making the technology of pedestrian navigation and positioning based on smartphones attract more and more attention.However,the low-cost smartphone IMU and the possible complex motion states of pedestrians undoubtedly bring huge challenges to the realization of high-precision pedestrian inertial navigation and positioning.To this end,this thesis uses the current high penetration rate of smart phones to carry out research on the indoor high-precision PDR positioning algorithm for pedestrians under complex motion conditions.The work in this thesis includes but is not limited to the following contents:Read a large number of documents related to indoor positioning technology,and summarize the current development status of pedestrian inertial navigation at home and abroad.At present,pedestrian inertial navigation technology is mainly divided into shoe-tie-based zero-speed update algorithm and waist-tie-based pedestrian dead reckoning algorithm.Since this thesis uses the smartphone with high penetration rate and rich built-in MEMS inertial navigation devices as the positioning platform,the research on the indoor inertial navigation positioning technology for pedestrians is mainly based on the waist-tied PDR algorithm.Aiming at the problems of low measurement accuracy and poor stability of the existing IMU devices built in smart phones,this thesis analyzes and calibrates the errors of IMU devices.The Allan variance method is used to analyze the random error of the gyroscope,the static data of the accelerometer is analyzed,and the calibration of the accelerometer is completed by the zero-bias calibration method.Use the ellipsoid fitting method to observe the magnetic interference phenomenon of the magnetometer and perform zero-bias calibrationAiming at the problem of imperfect classification of human motion state and insufficient motion state recognition accuracy in the current pedestrian inertial navigation technology,a human motion state recognition algorithm is proposed.The algorithm uses the moving average filtering method to perform data filtering processing on the MEMS collected data to reduce the large noise caused by the low-cost smartphone IMU devices,and then extract and analyze the common time-domain features of the acceleration data and the custom features mentioned in this thesis,and effective feature items are screened out.Then based on SVM multi-classifier and KNN classification to achieve high recognition rate of human motion state recognition,and synthesize sample sets of different sizes to compare the classification effect,and propose an adjacent gait constraint algorithm based on pedestrian motion law to achieve error classification The resulting motion state is effectively corrected,and finally high-precision recognition of the human motion state is realized.Aiming at the insufficient adaptability of the traditional PDR algorithm to pedestrian multi-motion states,an adaptive PDR algorithm under pedestrian multi-motion and an improved single peak detection method and zero-point search method are proposed to complete the high-precision gait detection for pedestrians in multi-motion states.Aiming at the limitation of gait detection for step frequency estimation,a high-precision step frequency detection algorithm based on dynamic time warping algorithm is proposed;for the problem that the step length is difficult to accurately estimate under pedestrian multi-motion,this thesis combines linear/non-linearity Step length estimation model and related digital features of single step under different motions of pedestrians,an adaptive step length estimation algorithm is proposed to achieve high-precision step length estimation;the fusion filtering of the heading angle estimation value is completed by the improved Kalman filter algorithm,And put forward the HDE algorithm based on pedestrian turning speed to complete the accurate correction of the heading angle.Improve the PDR position estimation method based on the result of the motion state recognition,and realize the high-precision navigation and positioning under the multi-motion state of pedestrians in the two-dimensional plane.