A Novel Bias-Compensated Linear Constrained Least Mean Squares Algorithm Over Distributed Network

In this paper, we propose a Diffusion Bias-Compensated Constrained Least Mean Squares (D-BC-CLMS) algorithm based on the idea of distributed estimation for adaptive filtering in network containing input noises. To reduce the interference of input noises, we use a new cost function. The variance of the input noises is derived by a novel method that uses some reasonable assumptions without any prior knowledge. Then we combine the diffusion strategy with BC-CLMS to improve the performance of single agent and to obtain more robustness. Eventually, simulation results confirm the theory is correct and demonstrate the excellent performance of the novel algorithm by comparing it with other conventional algorithms.