Unbiased blind adaptive channel identification and equalization. (Q2734278)

A low-cost algorithm for adaptive estimation of unbiased linear predictors in additive noise with arbitrary autocorrelation is presented in the paper. It is shown that the minimization of the prediction error variance, subject to a specific quadratic constraint, permits the derivation of an asymptotically noise-free prediction coefficient. The results are first obtained in the white noise situation and later extended to noise with arbitrary autocorrelation. A nice feature of the proposed technique is that it does not require the estimation of the noise variance, although some adjustment of the constraint value is needed. The global convergence is proven analytically. The performance of the proposed denoising algorithm is evaluated using data received over a simulated wireless test channel. The same method can be used for more general noisy regression problems, provided the regression variables are sufficiently correlated.