Structure and Adaptive Algorithm of Frequency Domain SSCF-ANC System

A gradient based active noise control system using the frequency domain SSCF (Square Sum of Correlation Function) adaptive algorithm without a model of the secondary path is proposed. The cost function of this adaptive system is a square sum of the cross-correlation coefficient of the input noise and the output error signal rather than mean square of the output error signal. A observable joint statistics of two signals instead of the estimation of a unobservable signal is considered. The spectrum detection method of the noise source and the influence of the detection error on adaptive algorithm are investigated, as a result the stability of the algorithm was proved. Since this algorithm is a normalized algorithm in the frequency domain, the adaptive parameters independently converge to the optimum value. Numerical verifications are also presented by using non-stationary noise source observed in passenger car and the actually measured transfer functions of the acoustic path.