An efficient low-complexity algorithm for crosstalk-resistant adaptive noise canceller
Résumé
This paper proposes an algorithmic and memory complexity reduction of the well-know crosstalk-resistant adaptive noise canceller (CTRANC). First, we show that the decorrelation approach is not accurate with the feedback structure. Consequently, a minimum based energy criterion approach more suitable with the crosscoupled structure, like the one proposed by Zinser et al., is proposed. Indeed, by using an additional stationarity assumption about the filter coefficients, we prove that recursive gradient formula could be rewritten as convolution operation. Simulation results show that the proposed algorithm performance are as good as the one proposed by Zinser despite its significantly lower complexity.