Signature Waveform Estimation By Using Short Training Sequences

Abstract

Multiuser detection techniques are known to be effective strategies to counter the presence of multiuser interference in code division multiple access channels. Generally, multiuser detectors can provide excellent performance only when the signature waveforms1 of all users are precisely known. Hence, the estimation of signature waveforms is a challenging issue in mobile communication systems. In this paper, we compare the performance of two training based estimators of using short training sequences. One is maximum likelihood type signature waveform estimator that requires the knowledge of spreading sequences and training sequences. The other estimator is based on subspace method and requires the knowledge of training sequences only. Through the simulations, we show the signature waveform estimation performance of both systems and the effect of the estimation error on the performance of a multiuser detector. The complexity comparisons of both systems are also given.