Research On Distributed Maximum Correntropy Criterion Adaptive Filtering Algorithm

Most distributed adaptive filtering algorithms are based on 2l norm optimization methods,which are not suitable for impulsive noise environments.Diffusion Affine Projection Sign Algorithm(DAPSA)based on 1l norm optimization method has strong robustness,but it lacks variable step size DAPSA at this stage.On the other hand,most distributed variable tap-length adaptive filtering algorithms build cost function of tap-length based on the 2l norm optimization method,which will fail to estimate the target quickly and accurately in impulsive noise environments.In order to solve the above problems,the variable step size DAPSA and distributed variable tap-length adaptive filtering algorithms in impulsive noise environment are studied in this thesis.Firstly,the adaptive filter and system identification model are briefly introduced.Then the adaptive network filtering,the common distributed adaptive filtering algorithm and distributed variable tap-length adaptive filtering algorithm are introduced.The convergence analysis of some distributed adaptive filtering algorithms is given in the thesis.Secondly,in order to solve the problem that DAPSA can't achieve faster convergence speed and lower steady-state error at the same time when the DAPSA converges with a fixed step-size in the impulse noise environment.A variable step-size method based on the maximum correntropy criterion and a Distributed Maximum Correntropy Criterion Variable Step Size Affine Projection Sign Algorithm(DMCCVSS-APSA)are proposed in the thesis,then the computational complexity of DMCCVSS-APSA is analyzed.Thanks to the improved chi-square kernel function,the proposed algorithm can achieve higher convergence speed and lower problem error at the same time.Simulation experiments show that DMCCVSS-APSA achieves better convergence performance than DAPSA in the impulsive noise environment.Finally,the existing distributed variable tap-length adaptive filtering algorithms based on the 2l norm optimization method are not suitable for the impulsive noise environment.The maximum correntropy criterion is used as the cost function of unknown vector weights and tap-length,then a Diffusion Variable Tap-length Maximum Correntropy Criterion(DVTMCC)algorithm is proposed.The convergence conditions of step size,transient behavior and computational complexity of the DVTMCC algorithm are analyzed in the thesis.Simulation results show that the DVTMCC algorithm can estimate the weights and tap-length of unknown vectors well in the impulsive noise environment.