Research On Spline Adaptive Filters And Networks

Adaptive filters have found widespread applications in interference cancellation,echo cancellation,active noise control,channel equalization,and so on.Most of the above applications can be characterized by system identification.In some applications,the desired response of the unknown system may suffer from nonlinear distortion.If a classical adaptive filter is used to identify such unknown system,it will yield large steady-state misalignment or even divergence.The spline adaptive filter proposed in recent years has provided good solution to the problem of nonlinear distortion.However,when the environment contains non-Gaussian noise,the performance of the spline adaptive filter will be degraded.In order to improve the performance of the spline adaptive filter in non-Gaussian noise environment,this thesis proposes a robust spline sign algorithm(RSSA)and a spline least mean fourth(SLMF)algorithm.The RSSA is robust against impulsive noise,while the SLMF can reduce the steady-state misalignment under uniform or binary noise.With the development of the network technology,adaptive filters have been widely used in distributed networks.In order to enhance the performance of distributed networks to estimate unknown parameter,this thesis extends the RSSA and the SLMF to distributed networks and proposed a diffusion RSSA(DRSSA)and a diffusion SLMF(DSLMF)algorithm.Finally,to give deeper insight into the stochastic behavior of the diffusion spline least mean square(DSLMS)algorithm,this thesis analyses the performance of this algorithm.