Connecting Steiglitz-McBride Identification, Active Noise Control, and Coefficient Filtering to a Common Framework

摘要: There are many adaptive algorithms that incoporate filtered versions of the regressor vector, error signal, parameter or update term. These modifications can often be viewed as attempts to recapture improve known stability properties linear LMS algorithms, despite operating with IIR parameterizations within a feedback loop. This chapter reviews some basic results on (exponential) and then applies them in three situations: Steiglitz-McBride family an analysis algorithm performance acoustic noise cancellation systems usually incorporate ‘filtered-u’ approach, consolidation coefficient filtering smoothing schemes such momentum leakage. The principle trading-off filters various signals is powerful unifying used create ‘new’ it allows translation between seemingly unrelated forms.