Parameter Correction Based on Particle Filter and Probabilistic Neural Network for Pedestrian Positioning

A particle filter was employed to fuse the measurements of GPS and wearable sensors,and a probabilistic neural network was adopted to recognize the pedestrian's activities according to the features extracted from the accelerations within fixed time windows.These methods reduce the errors of the azimuth,step length assessment and step counts,and correct the positioning parameters.Especially the real-time step length assessment by analyzing the vibration feature of accelerations and the sampling synchronization with time length proportional equivalence optimize the system model.The experimental results verify the effectiveness of the proposed approach.