| Adaptive filtering technology,due to its unique processing mechanism,is widely ap-plied in signal processing fields such as automatic control,target tracking and radar sys-tem.In the practical application of the adaptive filter,most noise environment of the sys-tem is non-Gaussian noise environment.Therefore,the design of adaptive filters is usually modeled on the background of non-Gaussian noise.At present,most of the adaptive fil-tering algorithms are based on the second-order criterion of error,but the performance of these algorithms in non-Gaussian noise environment is not very ideal.In this case,researchers have proposed algorithms based on non-second-order error statistics,where the generalized maximum correntropy criterion is a classical non-second-order statistical form.The aforementioned correntropy criterion contains error information of higher or-der terms,and can effectively suppress the adverse effects of impulse noise.In addition,similar to the real-valued system,the complex-valued system contains both Gaussian and non-Gaussian noise environments,and the real and imaginary parts of the complex signal can fully describe the basic characteristics of the system.Therefore,it is of great signif-icance to study adaptive filtering algorithms for complex systems.Complex least mean square algorithm is computationally efficient,but it is not robust to non-Gaussian noise while the generalized maximum complex correntropy criterion can effectively solve this problem.Hence,based on the generalized maximum complex correntropy criterion,a complex adaptive filtering algorithm for non-Gaussian noise environment is proposed in this thesis.The main contents are as follows:（1）Based on the generalized maximum complex correntropy criterion and combin-ing with affine projection,a new AP-type complex filtering algorithm,namely the affine projection generalized maximum complex correntropy criterion algorithm（APGMCCC）,is proposed.It is used for system identification in the impulsive noise environment.First-ly,the algorithm is to minimize the generalized complex correntropic loss function of the posterior error vector,and obtain the corresponding cost function after the l2norm con-straint on the weight vector.Secondly,the Lagrange multiplier method is used to trans-form the cost function into an unconstrained optimization function.Finally,the weight updating equation of the algorithm is obtained by calculating the gradient.（2）On the basis of research content（1）,the stability of APGMCCC algorithm is fur-ther analyzed,and the filtering performance of the algorithm is verified in complex Gaus-sian noise environment and complex non-Gaussian noise environment-impulsive noise.In addition,the filtering performance of this algorithm is compared with the generalized maximum complex correntropy criterion algorithm,results show that the former has better performance. |
