Theoretic analysis of the /spl gamma/-LMS algorithm

Abstract

The AR modeling is widely used in signal processing. The coefficients of AR model can be easily obtained by a LMS prediction error filter. However, it is known that such filter will give bias coefficients when the input signal is corrupted by noise. In previous works, Treicher [1979] suggested the /spl gamma/-LMS algorithm to reduce the bias problem caused by Gaussian noise. This paper gives the theoretical analysis of the /spl gamma/-LMS algorithm. We derive the close form solution of the second order statistics of the tap-weight vector. Computer simulations are provided to show the accuracy of our theoretical result.