抄録
As an adaptive algorithm for active noise control, LMS algorithm is most popular because of its simplicity and small calculation amount. However, LMS algorithm doesn't converge quickly and accurately because it needs high order of FIR model for colored noise, for example impact noise. For this problem, we propose a method using least squares lattice algorithm (LSL algorithm). LSL algorithm has high-speed convergence based on least-squares method, and accurate convergence by autoregressive moving-average (ARMA) estimation. Therefore, LSL algorithm is expected that it has quick and accurate convergence of ANC for impact noise. Also, we attempt to apply this LSL algorithm to a case close to actual environment. In this case, since the acoustic field has secondary path characteristic, the filtered-x algorithm is applied generally. However, filtered-x algorithm cannot apply to LSL algorithm directly because of it does not calculate FIR coefficients but ladder coefficients. Hence, We propose a method that do not use filtered-x algorithm, and show the effectiveness of that in computer simulations.
