Dynamic weight leakage for LMS adaptive linear predictors

Adaptive line enhancers provide a means of adaptively estimating the spectral characteristics of low SNR stationary signals. The LMS algorithm is a computationally efficient method for updating filter weights and is commonly used. Insufficient modal stimulation can lead to false spectral peaks. These peaks may be eliminated by applying fixed weight leakage however this produces a biased spectral estimate. Dynamic weight leakage and hybrid combined weight leakage schemes have been developed and are applied within the LMS structure. They provide mechanisms for suppressing state spectral modes with minimal or no bias in the spectral estimates. These schemes can be implemented with only a minor impact on the computational effort.