Research On Multi-Source Fusion Indoor Positioning Technology Based On Smartphone

In recent years,with the popularity of intelligent mobile devices,location-based services have risen rapidly,and accurate location information is particularly important.However,due to the serious attenuation of the GPS signal caused by the occlusion of the building,it cannot penetrate the building,and now there are high-rise buildings and high density everywhere,it is difficult for the indoor positioning technology to provide continuous and reliable services.In order to solve this problem,researchers have conducted extensive explorations.With the development of low-cost inertial sensor technology,smartphone-based indoor positioning methods have gained more and more researchers’ attention.Smartphone-based Pedestrian Dead Reckoning(PDR)can calculate the cadence,step length and heading by using the information collected by the built-in sensors of the smartphone,so as to solve the pedestrian track.However,the traditional PDR technology still has the problem of large error calculation limitation.When a pedestrian holds a mobile phone for complex activities such as running,jumping,and going up and down stairs,it will cause a large cumulative error in the step frequency,step length,and heading angle,resulting in low positioning accuracy.In response to the above problems,this thesis proposes a multi-source fusion indoor positioning method based on smartphones for human activity recognition assisted by PDR and geomagnetism,and realizes it on smartphones.The specific research contents are as follows.(1)Aiming at the current situation that traditional recognition methods cannot meet the research needs of human activity recognition technology,a deep learning model based on the combination of Wavelet Transform(WT)and Convolutional Neural Network(CNN)is proposed.The waveform data of the multi-channel sensor is decomposed and reconstructed by WT as input.The CNN with different convolution kernels is used to efficiently extract multi-dimensional features,and the maximum pooling layer is used to filter the interference noise caused by the unconscious shaking of the human body.After the output classification of the fully connected layer,the accurate identification of human activity status is realized.Experimental results show that the deep learning model based on the combination of wavelet transform and convolutional neural network achieves an F1 score of 91.65%,which has a higher activity recognition ability.(2)Aiming at the problem that the traditional pedestrian dead reckoning algorithm cannot adapt to the reliable positioning of the target in different motion states,a positioning method based on deep learning for human activity recognition assisted PDR is proposed.In the offline stage,a deep learning network is used to preprocess and train the built-in sensor data of the smartphone to obtain a human activity recognition model.In the online real-time positioning stage,the different motion states of the target are identified based on the model,and the traditional PDR algorithm is optimized from the two aspects of step count detection and step length estimation to realize real-time positioning under complex human activity states.The experimental results show that the positioning method can accurately identify a variety of complex human motion states,and the recognition accuracy is as high as 99.50%.At the same time,the accuracy rate of the advanced PDR step recognition algorithm is increased by 10.94%,and the maximum positioning error is reduced by about 16.2%,which verifies the effectiveness of the proposed algorithm.(3)Aiming at the problem that a single pedestrian dead reckoning method will produce large cumulative errors in the positioning process,a multi-source fusion indoor positioning method based on smartphone-based human activity recognition assisted PDR and geomagnetism is proposed.In the offline stage,the geomagnetic reference map database is established by using the geomagnetic signals collected by the built-in magnetometer in the smartphone.In the online real-time positioning stage,the geomagnetic positioning matching method based on the geomagnetic reference map database is realized.In the end,fusion positioning will be performed based on human activity recognition-assisted PDR method and geomagnetism,and real-time calculation and update of pedestrian position trajectory under complex human activity patterns on the smartphone side,realizing a multi-source fusion indoor positioning method based on smartphones.The experimental results show that the proposed multi-source fusion positioning method can accurately identify a variety of complex human motion states,reduce the cumulative error of the traditional PDR method,and the maximum error is controlled within 0.7 meters,achieving good positioning accuracy and having a stronger robustness.