A COMPARISON OF CYCLOSTATIONARY BLIND EQUALIZATION ALGORITHMS IN THE MOBILE RADIO ENVIRONMENT

In this paper we address the problem of blind equalization in the mobi le radio environment using cyclostationary techniques, where the chann el is generally non-minimum phase and the phase information of the sys tem has to be preserved. Conventional second-order statistics applied to the received signal sampled at the symbol rate do not preserve phas e information and therefore cannot be applied to identify separate min imum-phase and maximum-phase zeros of the channel. It has been shown t hat phase information can be preserved using only second-order statist ics if the received signal is sampled using fractionally spaced sampli ng. This paper deals with the issue of how the convergence of the cycl ostationary blind equalization algorithms compares to conventional sup ervised methods in the pan-European mobile radio system GSM. The resul ts suggest that the convergence speed of some of the actual cyclostati onary algorithms is fast for stationary channels but slow compared to conventional supervised non-adaptive (LS) and adaptive (LMS) algorithm s in time-varying mobile radio channels. (C) 1998 John Wiley & Sons, L td.