Unbiased LMS filtering in the presence of white measurement noise with unknown power (Q2732955)

The paper presents an algorithm for asymptotically unbiased estimation of the parameters of autoregressive models when the observations are contaminated by additive white noise. The authors propose to apply a first order filter to the observed process and use the resulting process to estimate the noise-induced bias in the estimators of the parameters. The estimated bias then can be subtracted from the standard estimates to achieve unbiasedness.