Parameter Estimation Of Phase Coded Signal Based On Compressed Sensing

The phase-coded signal has a simple modulation mode,flexible coding,strong anti-interference ability and low probability of interception.It is widely used in radar countermeasures and civilian communications.Estimating the carrier frequency and symbol rate of the phase-encoded signal is helpful for obtaining the signal modulation information.It becomes a research hotspot in the field of communication signal parameter estimation.Although the traditional cyclostationary analysis method can realize blind estimation,it requires a large amount of data and it is difficult to achieve real-time processing.Regarding the issue above,this thesis starts from the sparseness of the autocorrelation vector of the phase-encoded signal,and studies the parameter estimation of the phase-encoded signal based on compressed sensing.The major work is outlined as follow:1.A large number of data samples are required for traditional method of parameter estimation for phase-encoded signals.This thesis analyzes the sparseness of the signal in the cyclic autocorrelation domain.According to the compressed sensing theory,the cyclic autocorrelation vector is compressed,and the parameter estimation model based on compressed sensing of phase-coded signal is established.At last,simulation the compression and reconstruction process of cyclic autocorrelation vector.2.Combining the inherent structural features of phase-encoded signal,such as the sparseness of cyclic autocorrelation vector,symmetry,and the unity of the real and imaginary part support set,the signal is compressed and sampled,and a novel symbol rate estimation method based on the structure compressed sensing is proposed.Traditional sparse Bayesian learning methods cannot directly reconstruct the complex cyclic autocorrelation vector.This thesis transform the complex model into the real-valued one.From the analysis,it can be known that the latter part of the real-model observation component can be approximated by zero vector,which simplifies the numerical calculation,establishes a compressed sampling model to separate the real and imaginary parts,and reconstructs the two parts vector by a multi-task sparse Bayesian algorithm which shared a same support set.Experimental results show that this method can greatly reduce the number of measurements and improve the real-time performance of the algorithm.3.An adaptive hard threshold iteration method in impulse noise environment is proposed.In the compressed sensing processing,not only exists Gaussian noise,but also exists short-term large pulse interference.In this thesis,the Alpha stable noise model is established,and the influence of this kind of noise on the observation vector and the cyclic autocorrelation vector is analyzed.In the Alpha stable noise environment,the spectral lines of cyclic autocorrelation vector are completely submerged.However,since the observation vector is essentially a part of the delay product vector,the temporal sparse pulse noise does not destroy its periodic structure,and hard threshold can be used to preprocess the observation vector.The researche shows that the value of threshold has a great influence on the noise suppression performance of the algorithm.Under the condition that there is not enough prior information,this threshold is difficult to select.Therefore,in each iteration of signal reconstruction,the adaptive algorithm is used to select the threshold,and the cyclic autocorrelation vector can be recovered robustly to estimate the signal parameters.4.The Alpha-stable noise is approximately sparse in the time domain,and the traditional model constrained by the L2 norm failed.In this thesis,the properties of existing sparse norm are analyzed in detail,and the Lp norm,the Lorentzian norm and the correntropy induced metric are taken as examples to constrain the loss terms in the compressed sensing model.For the respective characteristics of these norms,the corresponding convex relaxation and smooth reconstruction algorithm is introduced.Experimental result shows that the proposed sparse norm model can effectively suppress the impulse noise and obtain robust parameter estimation of the cyclic autocorrelation vector.