Research On Distributed Diffusion Subband Filter Algorithms Based On Set-Membership

After decades of development,adaptive filtering which is an important part of digital signal processing,have been widely used in many fields.As the core content of adaptive filtering theory,adaptive filtering algorithms have always attracted the attention of scholars.To propose adaptive filtering algorithms with fast convergence speed,low steady-state error and low computational complexity is the goal pursued by researchers.A classic application scenario of adaptive filtering algorithms is system identification.Although traditional adaptive filtering algorithm such as NLMS algorithm is simple in structure and easy to implement,they can no longer meet actual needs when faced with a situation that input signal is highly correlated.The proposal of the subband adaptive filter(SAF)has become an excellent solution to such problem,by using its special structure can effectively improve the convergence performance of the adaptive filtering algorithms when the input signal is colored.In addition,the set-membership(SM)filtering technology which set a specified boundary value for error signal has also become an effective solution to improve the performance of the traditional adaptive filtering algorithms.Distributed filtering theory which combines distributed network and adaptive filtering theory,has been a popular research in recent years,and is widely used in many fields such as disaster relief management,target positioning and tracking,environmental monitoring,medical applications and so on.Distributed networks can be described as three strategies:diffusion strategy,incremental strategy,and probabilistic diffusion strategy which is based on the network topology,and the diffusion strategy has been widely studied.In a distributed network with a diffusion topology,each node can not only collect and process the collected information data,but also exchange information with all its neighbors.By doing so,the network can make full use of local estimation information,so as to achieve good global estimation performance.Based on the diffusion strategy,various diffusion distributed filter algorithms were proposed,but the existing algorithms still have some shortcomings and are worthy of further research for improvement.In order to improve the estimation performance of the diffusion distributed filter algorithms,this paper uses the advantages of the set-membership(SM)filtering technology and proposed the set-membership diffusion subband filter(SM-DNSAF)algorithm by setting a specified boundary of the error signal.This algorithm improves the update strategy of traditional algorithm.In distributed adaptive estimation,it can achieve faster convergence speed and lower steady-state misalignment,as well as reducing the computational complexity.Moreover,by using a smooth estimation method,a smooth estimation version of the set-membership diffusion subband filter(SSM-DNSAF)algorithm is proposed,which further improve the update process of the proposed algorithm and achieves better performance.This article studies the sparse systems,In this system,a large number of weight coefficients are close to zero or zero,which makes the performance of traditional adaptive algorithms worse.For this reason,researchers have proposed many adaptive filter algorithms based on the proportionate matrix strategy,which can effectively improve the convergence performance of algorithms in sparse systems.Based on the idea of proportionate matrix,this paper proposes two algorithms for sparse systems in distributed adaptive estimation,namely,the set-membership diffusion proportionate subband filter(SM-DPNSAF)algorithm and the smooth estimation version of the set-membership diffusion proportionate subband filter(SSM-DPNSAF)algorithm.The proposed algorithms speed up the convergence speed in sparse systems.