Channel estimation and signal detection for mobile communications over fading channels

Abstract

In this thesis, channel estimation problems for various mobile communication systems are investigated. As a result, two new channel estimation algorithms are proposed that mitigate the performance degradation due to the inadequacy of the existing algorithms for fast fading channels. The first algorithm, referred to as the PLLMS-CE, is based on adaptive linear estimation and outperforms the existing channel estimation algorithms in fast flat fading channels. For the adaptation, we introduce a new adaptive algorithm called the partial leaky least mean square (PLLMS) algorithm, which applies leakage only to the tap weight corresponding to the current pilot symbol. Through an analysis of the PLLMS adaptation process, we find that the tap weight vector updated by the PLLMS algorithm converges in the mean to the optimal Wiener solution for a leakage factor. The mean square identification error (MSIE) analysis is performed for the proposed and existing channel estimation algorithms. The analytical result shows that the proposed algorithm outperforms channel estimators based on adaptive linear prediction. To confirm the analytical result, computer simulations are performed. As an application example, CDMA-2000 reverse link is simulated for various channel estimation algorithms. The simulation results show that the performance gain of the proposed algorithm over the existing algorithms is about 0.6dB for 1% frame error rate (FER) when mobile speed is 100km/h. The second estimation algorithm is proposed for frequency selective fading channel. The algorithm is derived based on the assumption that fading is approximated by a polynomial function of time. Combined with the Viterbi equalizer and per-survivor processing (PSP), the proposed algorithm provides improved tracking property and thus eliminate the error floor which is observed when conventional channel estimators based on the standard recursive least squares (RLS) or LMS algorithm is employed. Simulation for the IS...