Research On Direct Spread Spectrum Sequence Signal Acquisition Based On Neural Network

Currently,there are three main communication methods: optical fiber,satellite and spread spectrum.By converting a narrow-band signal into a wide-band signal,the spreadspectrum communication technique processes strong anti-interference ability and has attracted wide attention.The acquisition of spread spectrum code is vital in the spread spectrum system.Traditional acquisition algorithms include time-domain acquisition and frequency-domain acquisition,etc.Despite the difference in algorithm design,these approaches all require large data amount and high computational complexities,thus resulting in considerable cost in piratical implementation.Due to the disadvantages of traditional acquisition,the thesis utilizes the theory of compressed sensing to construct the model of direct sequence spread spectrum(DSSS)signal.In this model,the sampling frequency is no longer limited by the Nyquist theorem,and hence the data amount can be significantly reduced.We take advantage of PN codes to construct orthogonal basis and introduces a parallel DSSS signal acquisition algorithm based on the alternating direction method of multipliers.The model can be solved in a distributed manner.Each sub-problem has a closed-form solution,which reduce the calculation amount.Secondly,in order to further improve capture efficiency and and alleviate the burden of resource consumption,we exploit neural network to solve DSSS signal acquisition.The neural network model is constructed based on four aspects: network structure,activation function,loss function and update method.The optimal neural network can be obtained by modifying the hyperparameters.The capture probability of the neural network is very close to that of the compressed sensing capture algorithm.Meanwhile,the neural network capture algorithm takes less time,which proves the feasibility and high efficiency of the algorithm.Finally,the FPGA implementation of DSSS signal capture is introduced in detail,including the compression sensing and neural network algorithms.After verification and comparison,the acquisition probability of the two acquisition algorithms on FPGA is the same as that of the algorithm simulations.The resource consumption of the compressed sensing algorithm is larger than that of the neural network algorithm.