改进的小波分解LMS算法在非线性系统辨识中的应用

对基于小波变换的自适应滤波技术中较为先进的D-LMS（Decomposition Least Mean Square）算法进行改进,推导出一种变步长D-LMS算法。通过建立非线性系统模型,在基于MATLAB的仿真实验中,分别得出原D-LMS算法和改进算法的系统辨识图形和数据。结果表明,两种小波分解自适应算法都能够很好的对非线性系统进行辨识,而改进的变步长D-LMS算法的收敛速度及跟踪速度更快,稳态误调噪声较小,即辨识结果更加精确。

The Application of Modified Wavelet Decomposition LMS Algorithm in Nonlinear System Identification

By developing the D-LMS(Decomposition Least Mean Square) algorithm,which is more advanced in the adaptive filtering technique based on wavelet decomposition,a novel changed-step D-LMS algorithm is deduced.More importantly,this research builds a nonlinear system identification simulation model,and by the MATLAB software,the figures and data of both algorithms are obtained.The experimental results show that both adaptive filtering algorithms based on wavelet decomposition can identify nonlinear systems well.Moreover,the changed-step D-LMS algorithm can get higher convergence rate,faster tracking speed,and lower steady misadjustment noise.In other words,more accurate data can be obtained by this novel algorithm.