Pedestrian Trajectory-Oriented Multi-Source Fusion Indoor Positioning Technology Research

With the rapid development of the Internet of Things and smart terminals,location services have created tremendous application value in many fields,such as road navigation,fire rescue,and medical management.Outdoors,the Global Navigation Satellite System(GNSS)provides high-precision location services for outdoor positioning and navigation,but indoors,satellite signals are severely blocked by buildings,resulting in low indoor positioning accuracy.Affected by this,various indoor positioning technologies have emerged.Although a single positioning technology has advantages,it also has some disadvantages.Among them,the wireless signal is restricted by the environment and hardware conditions,and the inertial positioning is limited by the accumulated error,which leads to a greatly reduced positioning accuracy.Therefore,the research on indoor multi-source information fusion positioning has gradually become the research focus.This paper takes the pedestrian trajectory as the research object,the smart phone as the carrier,uses the Wi Fi signal widely existing in the indoor environment and the built-in inertial measurement unit(IMU)information of the smart phone to propose and implement the extended Kalman filter under error analysis(Extended Kalman Filter(EKF)integrates Wi Fi and PDR positioning algorithms.The main work of this paper is as follows:(1)In the process of offline database building,the signal strength distribution received from the same access point(AP)at the same location is irregular,and the classification performance of a single classifier is poor during online positioning.This paper proposes to use enhanced Gaussian mixture The model(Enhanced Gauss Mixture Model,EGMM)establishes an RSS(Received Signal Strength,RSS)distribution model,and uses the integrated learning method of multi-classifier voting for real-time positioning.The number of sub-models is determined by Bayesian information criteria.In each sub-model,the expected maximum algorithm(Expectation Maximation,EM)is used to calculate the model parameters,and the multi-classifier ensemble model is trained in the sub-models.In the online positioning stage,according to the model The weight vector in the parameter assigns the weight to get the final position output.Experimental results show that the probability that the positioning error of the proposed method is less than1 m is 92.34%.(2)In the step detection,the traditional dynamic threshold and ordinary threshold algorithms cannot avoid the problem of false peaks and valleys.The three-axis acceleration data of pedestrian walking is analyzed,and a dual threshold binarization algorithm is proposed for step detection.Set a dual threshold,set the state quantity according to the acceleration modulus value meets different threshold conditions,and a state transition is recorded as one step.At different speeds of pedestrians,the accuracy of the algorithm in this paper is as high as 95%.Aiming at the problem of heading angle error accumulation,an adaptive fusion algorithm is proposed to suppress the heading angle error accumulation.The magnetometer can calculate the heading angle,but the accuracy is not high.Therefore,the accelerometer and the gyroscope are fused to obtain a heading angle.Then,according to the set magnetic induction intensity,the heading angle calculated by the magnetometer and the heading angle fusion are allocated.The weight of the time.The experimental results show that under the rectangular trajectory,the average heading angle error of the algorithm in this paper is 4.5°,which is lower than other algorithms.(3)Aiming at the advantages and disadvantages of Wi Fi location fingerprint positioning and pedestrian dead reckoning(Pedestrian Dead Reckoning,PDR)positioning technology,this paper proposes an EKF algorithm based on error analysis for fusion positioning.Through the analysis of the error source of the Constant Turning Rate and Velocity(CTRV)model,the error model in the prediction process is established and the uncertainty caused by the error is determined.Experimental results show that the average position error of the proposed algorithm under the S-shaped and Z-shaped trajectories is reduced to 1.396 m and 1.223 m,respectively,which is lower than the Kalman Filter(KF)and EKF.