CAJ | 학술논문

针对最小均方算法收敛过程中收敛速度与稳态误差的矛盾，提出一种基于反正切函数的归一化最小均方算法。该算法利用反正切函数和误差自相关的时间估计建立了步长与误差之间的非线性关系，抑制环境中的非相关噪声，同时引入归一化信号功率扩大输入信号的取值。仿真结果表明，该算法具有较快的收敛速度、较低的稳态误差，同时具备较好的系统时变跟踪能力。
침대최소균방산법수렴과정중수렴속도여은태오차적모순，제출일충기우반정절함수적귀일화최소균방산법。해산법이용반정절함수화오차자상관적시간고계건립료보장여오차지간적비선성관계，억제배경중적비상관조성，동시인입귀일화신호공솔확대수입신호적취치。방진결과표명，해산법구유교쾌적수렴속도、교저적은태오차，동시구비교호적계통시변근종능력。
Aiming at the contradiction between convergence rate and static error in convergence process of the least mean square ( LMS) algorithm, a normalised LMS algorithm based on arctangent function is proposed .In this algorithm, arctangent function and time estimation of the error autocorrelation are utilised to establish a non-linear relation between the step factor and the error to suppress the non-correlated noise in the environment; meanwhile, the normalised signal power is introduced to expand the value range of input signal .Simulation results confirm that the proposed algorithm has faster convergence rate , lower static error and has better capability of system time-varying tracking .