Active Noise Control Aimed Harmony Between Hummand and Environment. II. Recursive Algorithms for Active Noise Control.

L工ErikSSOn(Nels0n lndusαies諏c。)1. INTRODUCTION Active noise control as described in the now classic patent by Lueg utilizes a feedforward con trol structure"'.In this approach, the undesired noise is detected by an input microphone which then feeds a signal through an electronic filter to obtain the desired amplitude and phase relationship, after which a downstream loudspeaker is used to gener ate the cancelling sound wave, as shown in Fig. 1.This scheme exploits the peak in the cross correlation function between the signal at the input microphone and loudspeaker to obtain broadband attenuation.This peak occurs at a time associated with the time required for the acoustic wave to travel from the input microphone to the loud speaker.After cancellation, the peak in the cross correlation or, equivalently, the magnitude of the coherence between these two locations, has been significantly reduced.An early description of a digital approach to implementing this feedforward concept may be found in Kid0Z', while Burgess has described an adaptive version utilizing a finite impulse response (FIR) adaptive filter in the "filtered-X" algorithms".Due to the transfer functions, H, following the adaptive model, including the loudspeaker, the input, X, to the weight update calculation must be filtered by an estimate of these same transfer func tions, also designated H, to ensure convergence.Eriksson has extended these ideas to an approach that utilizes a recursive infinite impulse response (IIR) adaptive filter to automatically compensate for acoustic feedback, that has been referred to as