一种新的频域自适应算法

New frequency domain adaptive filtering algorithm

分析了现有自适应滤波算法,并且提出了在有色噪声背景下能够快速收敛的频域自适应新算法.使用牛顿法搜索性能表面和近似于递归最小二乘(RLS)算法的结构,利用现有的拟牛顿QN(Quasi-Newton)时域自适应算法原理,通过快速傅里叶变换(FFT)将其应用于频域.结合快速块最小均方自适应滤波算法FBLMS(Fast Block Least Mean Square)中的并行处理方法对算法的运算过程进行了改进.由于调整了数据格式和增益矩阵的系数加快了迭代过程的收敛,并且提高了信号处理的效率.附加的计算机仿真结果分别给出了在白噪声和有色噪声输入相同汉明窗条件下,新算法、LMS算法和拟牛顿算法QN的自适应系统辨识的效果比较图,表明新算法能有效用于色噪声下的自适应滤波.

After analyzing existing adaptive algorithms, a new frequency adaptive algorithm was proposed which can achieve faster convergence speed under the color noise background. It uses the Newton-type algorithm searching the performance surface and uses the similar structure of recursive least-squares （RLS） adaptive filtering algorithm. The time domain quasi-Newton adaptive algorithm to the frequency domain was applied by using fast fourier transform （FFT）, and the whole calculating duration according to the parallel method using in fast block least mean square（FBLMS） adaptive filtering algorithm. Because of changing the format of the input signal and the coefficients of the gain-matrix, the speed of the recursive procession convergence and the efficiency of the signal procession were increased. Additive computer simulation results present the comparison of the adaptive system identify effect after using LMS algorithm, QN algorithm and the new algorithm respectively, which is separately passed the same hamming window with the white noise and color noise inputs. The results indicate the new algorithm can be efficiently applied to adaptive filtering with color noise.