Reduction of computational burden of adaptively updating control

Abstract

An active control method and apparatus for controlling vibration or sound wherein the computational burden to adaptively update the control filter within the update model is reduced. The apparatus includes means for producing an input signal for input to the update model and the reduction block, means for generating an error sensor, and an output device for canceling sound and/or vibration at a selected location. The method uses a gradient descent algorithm such as the filtered-x LMS algorithm and replaces the long filter model with a shortened length filter model. Error signal information together with the output information from the shortened length filter model are used to update the coefficients of the control filter according to an update calculation method. The shortened filter model only contains frequency response information at the L frequencies of interest. Because of the reduced amount of information contained in the shortened filter model, more updates per unit of time can be performed on the control filter(s) or a smaller or fewer processors or CPU's may be utilized. The method has applicability to feed-forward and feed-back type systems and also to systems using a waveform generator for supplying the input signal. The reduction method may be implemented on-line or off-line.