Reduction of Gps Standard Receivers Noise Using Parallel-structure Wavelet Based Neural Networks

Position information obtained from standard GPS receivers is known to be corrupted with noise. To make effective use of GPS information in a navigation system it is essential to model this noise and to eliminate its effect. This paper present Parallel Structure Wavelet Based Neural Network (PSWNN) for predicting the Differential GPS (DGPS) corrections. The PSWNN consists of multiple numbers of WNNs connected in parallel. Each WNN in the PSWNN predicts the same DGPS corrections future value based on input data with different embedding dimension and time delay. The embedding dimension is chosen optimally to have superior performance for each time delay value. The PSWNN determines the final predicted value by averaging the outputs of each WNN. The performance of proposed PSWNN is compared with WNN in application of DGPS corrections prediction. The proposed algorithms in DGPS system are implemented by a low cost commercial Coarse/Acquisition (C/A) code GPS module. The experimental results demonstrate which the PSWNN has great approximation ability and suitability in DGPS connection prediction than WNN; so that, the PSWNN prediction accuracy respect to the WNN is improved from 1.7094 to 0.9889 meters for 10 seconds prediction and from 2.2652 to 1.8352 meters for 30 second prediction, respectively.