Blind channel estimation without eigendecomposition
Özgür Özdemir, Murat Torlak · 2004
We address the problem of blind estimation of finite-impulse response (FIR) channels without eigendecomposition in CDMA systems. The blind channel estimation problem arises in the presence of frequency selective fading, when the training data is unavailable. The blind channel estimation techniques in the literature often rely on the eigendecomposition of the received signals. The blind channel estimation can also be coupled with the multiuser receiver design in the context of minimum output energy (MOE) criterion. While MOE receiver avoids eigendecomposition, the inverse of data covariance matrix of received signals is needed [Z. Xu et al., Nov. 2002]. In this paper we propose a blind channel estimation method that avoids eigendecomposition. At the same time, the proposed method directly uses the data covariance matrix without requiring its inversion. Analytical and simulation results are provided to show that the channel estimates converge to the eigendecomposition based channel estimates at the medium to high signal-to-noise ratio (SNR).