一种新的变步长NLMP快速算法

A New Variable Step-Size NLMP Fast Algorithm

由于NLMP算法的收敛速度不够快,它不能应用于实时性要求高的场合。引入指数遗忘因子加权误差的P阶相关,用数据段的P阶相关的平均更新步长,每L点数据段更新一次步长,提出了一种新的变步长NLMP算法。算法仿真结果表明,新的算法保持了传统NLMP算法的各项优点,同时在稳定性和收敛速度方面比传统的NLMP算法都有很大的改善,便于对信号进行实时性处理,算法简单易于实现,对高斯噪声和α稳定分布下的尖峰脉冲噪声都有很强的韧性。

Since the convergence speed of NLMP algorithm is not very quick, it can not be used in some situations where the demand of real - time processing is very strict. Making use of the forgetting factor to weigh the p-rank correlation of the error and the average of the p - rank correlation of the data block to update the step - size, and every L point data block updateing step - size once, the paper proposes a new modified NLMP algorithm - - variable step - size NLMP fast algorithm. The simulation results indicate that this new algorithm inherits all the advantages of the NLMP algorithm and both convergence speed and stability of the new algorithm are better than that of the traditional NLMP algorithm, and it is convenient for real -time signal processing, meanwhile, this algorithm is very simple and easy to realize. It is robust to both gauss noise and pulse noise in α -stable distribution.