摘要
在数字音频里,由于LMS算法具有低计算复杂度、在平稳环境中的收敛性好和利用有限精度实现算法时的稳定性等特性,使LMS算法成为自适应算法中应用最广泛的算法。本文对LMS算法及其改进算法进行了研究,探讨了步长因子μ(n)对各种算法收敛性、稳定性的影响。结果表明,变步长μ(n)的取值尤为重要,如果μ(n)取较大值则具有较快的收敛速度,如果μ(n)取值很小,则NLMS算法近似等效于LMS算法。它们的自适应过程较快,性能有了很大改进。
In digital audio, the LMS algorithm becomes the most widely used algorithm of the adaptive algorithms with its low computational complexity, good convergence characteristics and stability when used in finite precision algorithm in a stationary environment. This paper studies the LMS algorithm and its improved algorithm, explores the impact of step factor u （n） to the convergence and stability of various algorithms. The results show that the value of u （n） is particularly important. The convergence speed will be fast if the value of un） is large; otherwise if the values of u（n） is very small, the NLMS algorithm is approximately equivalent to LMS algorithm. The adaptive process is fast, and the performance has been axeatlv improved.
出处
《电脑与电信》
2013年第4期60-62,共3页
Computer & Telecommunication
