基于箕舌线函数的变步长归一化最小均方算法

对变步长归一化最小均方(VS-NLMS)自适应算法进行了讨论,针对其在自适应过程渐进稳态时对噪声干扰过于敏感的不足做了改进。同时,为了协调其低稳态误差与快速跟踪性能间的矛盾,引入基于相关误差项的变步长调整方案,同时采取了替代Sigmoid函数的箕舌线函数作为步长迭代公式,大大降低了计算复杂度。仿真结果表明,改进后的算法不仅具备优于归一化最小均方算法的收敛性能,同时具备了更小的稳态失调和快速灵敏的时变跟踪能力。

Modified Variable Step - size Normalized Least - mean - square Algorithm Based on Versoria Function

Variable Step-size Normalized Least-Mean-Square(VS-NLMS) algorithm is discussed,it is improved that sensitive to noise disturbance when it comes into steady statement in adaptive process.Moreover,to coordinate the conflicting requirement of low misadjustment and fast tracking rate,step size of the filters is adjusted according to the square of the time-averaged estimatinon of the autocorrelation of the noise signal.Meanwhile,versoria function is used in the new algorithm instead of the Sigmoid function for the modified complexity of the calculation.Furthermore,some simulation examples in applications of noise cancellation show the validity of the new VS-NLMS algorithm compared with other modified adaptive filtering algorithms.It has smaller steady disorder and fast time variable tracking.