New frequency domain adaptive filtering algorithm

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 ofthe 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.