离散分数余弦变换自适应滤波算法及应用

An Adaptive Filtering Algorithm of Discrete Fractional Cosine Transform and Its Application

为改善时域最小均方算法的收敛速度,提出离散分数余弦变换自适应滤波算法。利用离散余弦变换的去相关能力,构造一个新型自适应算法所需的目标函数,采用递归方式使变换域信号功率归一化,选取最优的变换阶次,获得最小均方误差,明显改善了算法的收敛速度。仿真实验结果显示:与传统的LMS、DCT-LMS相比,离散分数余弦LMS算法的特征值比有比较明显的下降;DFRCT-LMS比前两种算法更快地收敛。在语音增强应用实验中,离散分数余弦LMS算法优势明显,具有实际应用价值。

The adaptive filtering algorithm of discrete fractional cosine transfrom is proposed in order to increase the convergence speed of LMS algorithm. Because of superlatively decor - relating effect of discrete fractional cosine transform, the proposed algorithm with the optimal fractional order, forms objective function within transform output power is normalized by recursive form. The minimization of mean square error implies that the algorithm can remark- ably increase the convergence speed. The simulation results indicate that, compared with those of LMS and DCT - LMS, the EVR of DFRCT - LMS is remarkably reduced; the DFRCT - LMS method can speed convergence effectively. DFRCT - LMS also behaves well in the application of speech enhancement,and has wide application value .