A time-varying normalized step-size based generalized fractional moment adaptive algorithm and its application to ANC of impulsive sources

This paper develops a generalized fractional low-order moment-based adaptive filtering algorithm well suited for active noise control (ANC) of noise sources exhibiting peaky characteristics. In a previous algorithm, the impulsive nature of the noise sources has been addressed by employing a normalized step-size and the fractional power of the error and reference signals. In this paper, we employ a convex combination approach to develop a variable step-size (VSS) strategy. Thanks to the VSS in the proposed algorithm, the ANC system achieves a fast convergence speed as well as an improved steady-state noise reduction performance. Extensive simulations have been carried out for various scenarios, which show that the proposed algorithm is robust against the impulsiveness of the noise source as well as for a possible non-stationarity in the acoustic environment.