효율적인 가변 변수를 이용한 아크탄젠트 비용함수 기반diffusion NLMS 알고리즘

초록

In this paper, we propose an arctangent-cost-function based diffusion NLMS algorithm with an efficient variable parameter, robust against impulsive noise in the adaptive network. Compared with conventional algorithms where the variable parameter is an exponential function and the occurrence probability of impulsive noise is required, the proposed distributed estimation algorithm, utilizing the characteristics of the arctangent cost function is of less computational complexity and requiring no prior knowledge on impulsive noise, yields improved MSD convergence. Finally, its convergence performance is verified by using computer simulations for system identification.

키워드