Research And Application Of Indoor PDR And Signal Fingerprint Fusion Positioning Method

With the popularity of mobile smart devices,people’s demand for location-based services has increased.The micro inertial sensors embedded in the device and supported wireless communication protocols have led to extensive research and application of Pedestrian Dead Reckoning(PDR)and fingerprint positioning techniques based on received signal strength.The positioning continuity of PDR is good,and the model is simple and easy to implement,but errors cannot be avoided in each positioning,and long-term use will cause a large error accumulation.Meanwhile,there are problems with the adaptation of positioning models and inertial data sources in different walking modes.Fingerprint positioning has high accuracy,and hardware deployment is relatively simple.However,the online matching stage exists mismatch due to electromagnetic interference,multipath effect,human body and wall occlusion,resulting in increased positioning errors.Furthermore,wireless signals have time-varying characteristics,which will lead to unstable positioning results.To solve the above problems,this paper studies indoor positioning technology based on PDR and signal fingerprints.The main work and innovation points of the paper are as follows:(1)To address the problem that traditional PDR is only applicable to a single walking mode,which can result in low positioning accuracy or unusable algorithms in different walking modes,a PDR method that considers walking modes is proposed.After collecting data from the device’s built-in sensors,a soft interval Support Vector Machine multi-classification model is used to identify different walking modes by analyzing gait patterns and selecting and extracting multi-dimensional statistical features based on sliding windows.Appropriate data sources for each mode are selected by analyzing inertial data.An adaptive model is introduced,which dynamically adjusts the peak and valley threshold based on peak gait detection,to match the variable walking speed,improving the adaptability of the detection process to changes in motion state.To achieve heading estimation in multiple modes,the device’s attitude is calculated based on quaternion or Principal Component Analysis.Experimental analysis shows that the proposed algorithm can accurately identify different walking modes,and its positioning cumulative error is smaller than traditional PDR.(2)To solve the problems of cumulative errors in PDR and fluctuations in fingerprint positioning,a fusion positioning method is proposed that combines the strengths of both techniques.This method takes into account the feasibility of wireless signals for fingerprint positioning and corrects the weights of similar points in the signal space using motion spatiotemporal information.The corrected weights are then summed to improve the accuracy of fingerprint positioning and overcome environmental interference.The proposed method uses an Extended Kalman Filter(EKF)to linearize the fusion system model and achieve stable and accurate optimization results through the iterative prediction and update process of the positioning points.Experiments have verified that the proposed method can improve the accuracy and robustness of the positioning system when compared with a single PDR or signal fingerprint.Moreover,when compared with other fusion positioning methods,EKF performs better in terms of real-time positioning requirements and positioning accuracy.(3)The paper describes the design and development of both the front-end client and backend server of a PDR and signal fingerprint fusion localization system,along with the deployment and application of the proposed method.The Android phone sensor framework and Bluetooth low energy module are used to collect and store the sensor data and signal strength in the local database.The data is then uploaded through the network to create an offline fingerprint database and inertial database for online matching and classification model training.The sampled data is uploaded in real-time,and the destination coordinates are manually uploaded,which are then calculated by the back-end to enable interactive display of positioning and navigation on the front-end.The test results show that the system implemented in this paper basically meets the needs of indoor positioning.