성능 보장 분석을 통한 적응 필터 알고리즘 설계

Abstract

This works presents two adaptive filte ralgorithm ,schedule dstepsize sfo rth enor-malized least-mean-squares (SS-NLMS) and an Affi neProjecti onAlgorit hmBasedon Reuse Time of Input Vectors.The SS-NLMS algorithm uses pre-calculated stepsize cames from guaranteedperformance analysis with establishing an objective curve. This algorithms showsnot only good performance but also robustness with respect to differen tsignal-to-noise ratio (SNR) in measurement noise and differen tcorrelatio ni ninpu tsignalswith a very small number of on-line computations. Moreover, the scalability of thetabled stepsize with respect to the number of taps is described. For the efficientmemory usage in practice, a modifie dversio nreplace sth etable dstepsize sb ydown-sampled stepsizes with no performance degradation.The an APA based on the concept of reuse time of the current input vector isdeveloped from the concept of reuse time that is define da sth esurviva lperio do faninput vector, during which the input vector is continuously reused in the subsequentupdate equations. The algorithm consists of two key procedures: assignment andreduction. The assignment procedure assigns a fundamental reuse-time that is de-termined by the statistical criterion. Based on the guaranteed performance analysis,the algorithm determined whether the current error reaches the steady-state value,which indirectly controls the number of input vectors

this leads to fast convergence and small estimation errors. Through these two procedures, the proposed algorithmachieves not only improved performance but also extremely low computational com-plexity.