Combined Method for Adaptive Channel Estimation with Application to DMT Systems

Abstract

We propose a Second-Order Statistics (SOS)-based channel estimator that finds application in Discrete MultiTone (DMT) systems. Most SOS adaptive channel estimators are based either on Newton–Raphson or on steepest descent methods. These classes of methods have complementary advantages and disadvantages. In this contribution, a new adaptive channel estimator based on combined Conjugate Gradient-NR method is developed. Among a number of attractive features, the combined estimator demands less computational complexity than that required by NR-based estimators, and when compared to CG estimators, the combined estimator exhibits better convergence especially for ill-conditioned channels. All mentioned theoretical results are illustrated by numerical results for estimates of randomly generated and real measured channels. Comparison with other well-known algorithms shows good trade off between performance and complexity.