An improved filtered-x least mean square algorithm for cancellation of single-tone and multitone noise

The notable advantages of Filtered-x Least Mean Square (FXLMS) Algorithm over other algorithms in Active Noise Control (ANC) environment are its simple time realization, Low computational complexity, robustness and simplicity though the algorithm has slow convergence. In order to optimize its frequency dependent convergence value, a convergence parameter is used. Thus upon application of this algorithm to noise cancellation techniques especially when the noise contains multiple tones, the optimization of the convergence parameter can only be done for one of those tones. The rest of the tones will have a convergence rate slower than the tone for which the convergence parameter is optimized. Again the remaining tones will have less attenuation than they would have if they were to be optimized separately for different tones. In this paper, two more filters are added to the existing structure of the FXLMS algorithm. The concept of Soft Threshold is also introduced in order to achieve better performance in terms of speed of convergence. Furthermore the proposed algorithm is applied to noise cancellation having both Single tone and Multi tone noise and the results are compared with the existing FXLMS and Filtered-X Recursive Least Square algorithm.