| The breakthrough of new technology supports the rapid upgrade and regeneration of wireless communication,and the breakthrough of technology bottleneck depends on the innovation of related basic theories.Chaos has very excellent properties,such as easy generation,initial value sensitivity,randomness,wideband and consistency with communication requirements,thus the researchers are devoted to exploring the application of chaos in communication.Chaotic spread spectrum and chaotic carrier have become IEEE local wireless network communication standards,IEEE Std.802.15.4a and 802.15.6.In recent years,some new properties of chaos have been reported,such as the Lyapunov exponents spectra invariance,which makes chaos can be applied to baseband communication to improve the performance of traditional digital baseband communication.In theory,chaotic baseband wireless communication system(CBWCS)can not be affected by multipath transmission completely,but in engineering,it cannot achieve the best performance because of the need for future symbols to decode.The rapid development of Artificial Intelligence(AI)technology in the field of communication has achieved a lot of achievements,how to use AI technology to improve the performance of chaotic baseband wireless communication,has important theoretical significance and practical value.This disseration mainly studies the application of AI to improve the performance of CBWCS,including chaotic baseband waveform prediction and symbol information prediction,optimal decoding threshold prediction,direct information decoding.The autocorrelation function(ACF)invariance of chaotic baseband signal is found in the research process,and the application of this finding in blind wireless channel identification and 5G communication is presented.The main work is as follows:In CBWCS,the chaotic baseband signal generated by the chaotic shape-forming filter(CSF)can theorectically analysis the inter-symbol interference(ISI)caused by multipath transmission,which can be fully eliminated by the optimal decoding threshold in theory.However,the calculation of the optimal decoding threshold not only needs the wireless channel information,the past transmitted symbol information,but also the symbol information untransmitted.In practical communication process,the optimal decoding threshold can not be obtained because the symbol information untransmitted is not available at the current time.Aiming at the difficulty of obtaining the theoretical optimal decoding threshold in CBWCS,predicting the chaotic baseband waveform to obtain one future symbol information prediction based on echo state network(ESN)is proposed,so as to obtain more accurate decoding threshold.Predicting the future symbols directly based on convolutional neural network(CNN)to calculate the decoding threshold is proposed.Predicting the decoding threshold directly based on ESN to further improve the decoding threshold accuracy and the bit error rate(BER)performance of the chaotic baseband communication.The wireless communication BER performance can be improved by obtaining the decoding threshold which is much closer to the theoretical one,but the chaotic waveform is mapped into one dimensional threshold,thus the information contained in the waveform which contributes to accurate decoding is compressed and lost.A direct information decoding(classification)method based on support vector machine(SVM)is proposed by converting the symbol decoding problem based on threshold into a classification problem based on the chaotic waverform.From a new viewpoint,the wireless communication decoding process is converted into a pattern recognition problem,in which the received chaotic signal at the receiver is classified directly.Thus,more available information is preserved by high dimensional spatial mapping,and better performance is expected to achieve.In order to reduce the structural risk of SVM,genetic algorithm(GA)is used to optimize the SVM structure,then the discriminant function is obtained to decode the symbol information directly.In this method,the wireless communication decoding process is greatly simplified as compared to the methods using decoding threshold,moreover,it avoids extra error caused by compicated channel estimation and threshold calculation,which makes the potential of chaotic baseband signal for symbol decoding to a limit.Simulation and experiment demonstrate that the BER performance obtained by the proposed GA-SVM method is better than the threshold decoding methods,and much closer to the theoretical optimal one.The autocorrelation function(ACF)invariance of the chaotic baseband signal is found and demonstrated theoretically in the study,that is,no matter what data information is encoded by the chaotic baseband signal generated by CSF,the ACF of the chaotic baseband signal is always consistent with the that of the base function of CSF,which is referred to as the chaotic signal ACF invariance.This fundamental finding of the chaos property is used for blind identification of wireless communication channel,which can effectively simplify the calculation process and decrease the difficulty of channel identification.The analytical relation among the ACF of received signal,the ACF of the transmitting signal and the channel parameters is illuminated,the channel parameters can be solved directly by calculating the ACF of received signal at the receiver according to this analytical relation.Since the autocorrelation operation is robust to noise,the blind channel identification based on the chaotic baseband communication has good identification accuracy in the case of low signal-to-noise ratio(SNR).The superiority of blind identification based on chaotic signal is verified by simulation.To solve the problem that low delay resolution and noise still affects the identification accuracy of the blind channel identification analytical equation based on chaotic signal ACF invariance in practical application,the channel identification method based on deep neural network(DNN)is proposed.In this method,the DNN with stacked denoising autoencoder(SDAE)is used to learn the mapping relation between the ACF and channel parameters.On the one hand,the noise effect is further reduced by the denoising autoencoder(DAE),on the other hand,the delay resolution is further impoved by the generalization ability of DNN,and the superiority of this method is verified by simulation.The chao-based non-orthogonal multiple access(NOMA)is proposed by combining the chaotic baseband wireless communication with the NOMA technology in 5G communication.On the one hand,the chaotic signal generated by CSF is used as transmission information,and the corresponding matched filter(MF)is adopted at the receiver to reduce the noise effect.On the other hand,the aforementioned DNN with SDAE structure is used to identify the Rayleigh fading channel more accurately for improving the accuracy of power allocation and the decoding effect of power domain non-orthogonal.In addition,the proposed DNN network uses the off-line traing and online prediction mechanism,which addresses the problem of high difficulty and low accuracy in the identification of the rapid changing channel parameters for the current 5G NOMA technology.Simulation results demonstrate that the BER performance of chaos-based NOMA system is significantly improved as compared to the classical binary phase shift keying(BPSK)method. |
PDF Full Text Request