| With the increasing complexity of modern communication environment,it becomes more and more difficult to obtain useful signals on the premise of unknown the prior information of transmission channel and source signals.The proposal of blind source separation technology provides the possibility to solve this problem.In practical communication scenario,multi-channel signals could only be received by a single sensor and recovered according to the mixed signal due to the limitation of environment,equipment,cost,and other factors,which makes the single channel blind source separation algorithm have important research value and wide application prospects.This thesis focuses on the single channel blind source separation technology of communication signals from three aspects: virtual multi-channel,state estimation and energy difference.The main research contents are as follows:(1)A single channel blind source separation algorithm based on adaptive variational modal decomposition is proposed.Firstly,in view of the problems that the number of decomposition layers cannot be determined adaptively and the number of iterations is generally too high in the variational modal decomposition algorithm,an adaptive variational modal decomposition algorithm is proposed.According to the different noise levels in mixed signals,different spectral correlation coefficient thresholds are correspondingly set to achieve the effect of determining the number of decomposition layers under different signal-to-noise ratios adaptively.Secondly,an adaptive source number estimation algorithm is proposed to estimate the number of sources of mixed noisy signals under different signal-to-noise ratios accurately.Then,the decomposed components and the observed signals are combined to form a virtual multi-channel signal,the number of source signals is estimated,the average value is removed and whitening is performed.Finally,the classical blind source separation algorithm is used to separate the multi-channel signals,and the estimation of the source signals are obtained.The simulation results show that this algorithm can determine the number of decomposition layers and source signals of the mixed signal under different signal-tonoise ratios and different aliasing degrees adaptively.While separating the source signals from the mixed signal accurately,it can improve the adaptability of the algorithm,avoid redundant iterations and reduce the complexity.(2)A single channel blind source separation algorithm based on parameter estimation and Kalman filter is proposed.Firstly,in view of the limitations of Root-MUSIC(Multiple Signal Classification)algorithm in the estimation of similar carrier frequencies,an adaptive RootMUSIC algorithm is proposed.According to the conjugate symmetry relationship between the roots of polynomials,when selecting the roots closest to the unit circle,judge whether the normalized frequency corresponding to each root is equal to avoid selecting the roots from the same signal source,so as to improve the accuracy of carrier frequency estimation and obtain the estimation of the number of source signals.Secondly,the mixed signal is demodulated coherently,and the phase estimation of the source signal is obtained by using the corresponding relationship between baseband symbols and phase.Then,the signal model is constructed according to the obtained parameter estimation.Finally,the signal model is used as the observation vector of Kalman filter system,and the two processes of “time update”and“measurement update”are performed to obtain the best estimation of the source signals.The simulation results show that this algorithm can realize accurate single channel blind source separation for spectrum aliasing signals with different modulation modes,and the estimated signals can not only ensure strong correlation with the source signals,but also recover the time-frequency characteristics of the source signal well.(3)A single channel blind source separation algorithm for mixed signals with energy difference is proposed.Firstly,the energy ratio between source signals and the frequency range of each source signal in the mixed signal is estimated.Secondly,aiming at the problem of how to fully decompose the mixed signals with energy difference,a variational modal decomposition algorithm based on energy differences is proposed,which sets the dual thresholds of spectral correlation coefficient and central frequency.By comparing the correlation between each modal component and the mixed signal and the frequency range where the center frequency is located,the source of each modal component is identified,and the modal components from the same source signal are combined to obtain the component corresponding to the source signal.Then,the obtained signal components and the mixed signal are combined to form a multi-channel signal,and the signal is de averaged.Finally,the classical blind source separation algorithm is used to separate the multi-channel signals to obtain the estimated signals.The simulation results show that when the signal-to-noise ratio meets certain conditions,the algorithm can separate the source signals with different energy differences from the mixed signals with spectrum aliasing effectively,and the estimated signals can recover the symbol information of the source signals accurately.In summary,the problem of single channel blind source separation of communication signals is studied from three aspects in this thesis,and the effectiveness and feasibility of these algorithms are proved by simulation experiments. |
