ABSTRACT
A bstract System identification plays a central role in control applications and in a similar way also in active noise control. In active noise control the filtered-x type algorithm has mostly been used. Here, information about the secondary path and feedback path are needed to updatethe controller. Most acoustic systems can be described by an impulse response that can be measured by correlation methods (e.g. by an FFT analyzer) or identified by an adaptive digital filter. The identification by adaptive filter has some advantages concerning convergence speed and on-line capability. This paper deals with the adaptive LMS algorithm for identification. The first part describes the implementation of the adaptive filter and identification results. A good agreement between identified and measured results is obtained after a careful adjustment of the measurement bias. In order to achieve an appropriate filter order, the residual error was investigated and the BIC criterion is used. In the second part of the paper an improvement is suggested based upon the idea that the a priori information about the known parts of a system, in our case the low pass antialiasing filter and the reconstruction filter, can be utilized to reduce system complexity. This results in a faster convergence of the adaptive filter. Simulation and experimental results verify the improvement.
