Optimizing the computational effort of satellite signal acquisition based on mean recognition convolutional neural network

Abstract

Abstract The ability to synchronize the satellite signals affects the navigation performance. Only after a procedure of acquisition, namely coarse synchronization, the data information needed to position can be extracted from the satellite signals. Conventional acquisition methods rely on extending the signal integration time to improve the correlation quality of the signal. However, extending the coherent integration time results in a large computational burden, which is unrealistic for civilian receivers. In this paper, we try to optimize the computational effort in the satellite navigation signal acquisition phase and propose an algorithm based on the mean value recognition convolutional neural network. The matrix with a high mean value is separated by the convolutional neural network, and then the Doppler shift and code phase from the matrix with a high mean value is introduced, which can effectively change the computation to 1/4338 of the original one. The trained model of the neural network can be embedded into the receiver directly, which has important engineering applications.