Denoising of GPS Positioning Data Using Wavelet-Based Hidden Markov Tree

Precise position and navigation with GPS is always required for both civil and military applications. The errors and biases associated with navigation will change the positional information from centimeters to several meters. To estimate and mitigate the errors in GPS positioning data, the wavelet transform is most significant technique and proven. The traditional wavelet threshold methods will work to a certain extent but are not useful to estimate the signal levels to the expected level due to their incapability for capturing the joint statistics of the wavelet coefficients. The wavelet-based hidden Markov tree (WHMT) is designed to capture such dependencies by modeling the statistical properties of the wavelet coefficients as well. In this paper, a WHMT is proposed to reduce positioning error of the GPS data. To establish proposed method, the position data are decomposed using wavelets. The obtained wavelet coefficients are subjected to Discrete Wavelet Transform (DWT) as well-proposed WHMT for noise removal. In this proposed methodology, an Expectation Maximization (EM) algorithm used for computing the model parameters. The root-mean square error (RMSE) of proposed method shows better performance comparatively classical DWT.