


Home > Publication > Ph.D. Dissertation 









Title 

Active adaptive noise control for systems with cancellation path 


Author 

H.S.Kim 


Type 

KAIST Ph.D. Dissertation 


Year of Pub. 

1998 


We must consider 'cancellation path' when we deal with An adaptive algorithm in ANC system, otherwise algorithm become unstable. Existence of the cancellation path requires heavy computational power to run the adaptive algorithm especially when it is timevarying or of MIMOtype. In such situations, the algorithm has to be complex and suffer from slow convergence speed. There are three major objectives for the ANC algorithms to be developed: (a) computational effectiveness; (b) robustness to the environmental variations; (c) and fast attenuation rate of the residual error. The computational effectiveness will be pursued by simplifying the cancellation path model or by omitting some part of the algorithm and decoupling the MIMOtype cancellation path. Robustness to the environmental variations will be achieved by online cancellation path modeling technique so that the algorithm can adjust to the change in the cancellation path. Fast convergence can be achieved by using a modified error and/or normalization technique. It also can be accomplished by decoupling the MIMO cancellation path. Chapter I introduces the ANC system with historical review, objective, and scope of the dissertation. Successive three chapters describe three different but related ways to achieve specific objectives. In Chapter II, to reduce the amount of computations while accepting some performance degradation, an approach intentionally simplifying the cancellation path is proposed. The steady state weight vector solution in the presence of model error is investigated to show the validity of using simplified cancellation path model. DXLMS algorithm, being based on the hypothesis that the cancellation path model can be represented by a delay, is proven to be useful for narrowbandnoise cancellation application. In Chapter III, the shortcomings of the conventional online approaches such as the biased estimation of the cancellation path model and the performance deterioration due to error in the cancellation path model are addressed. In order to overcome the aforementioned shortcomings, introducing a unifiederror and injecting random probing signal are tried. The unifiederror includes the control error and model error at the same time and the magnitude of random probing signal is adjusted in accordance with the estimation of the model error. In Chapter IV, a method that converts the MIMO cancellation path into decoupled multiple SISO systems, is proposed. For each decoupled SISO cancellation path, CDXLMS algorithm and its normalized version are proposed to achieve faster convergence speed. In the decoupled system, the convergence speed of each mode can be assigned identical, which was impossible for conventional MIMO systems, therefore, convergence speed can be much improved. Moreover, great reduction of programming complexity and computational burden are possible in realtime adaptive control implementation. Chapter V concludes and summarizes the proposed works.








