Cyclostationary signal processing in multi-antenna systems: From direction-of-arrival estimation to channel estimation and equalisation

Cyclostationarity is a statistical property present in most communication signals and it is measured in terms of the spectral correlation exhibited by the signal. Communication signals exhibit spectral correlation at harmonics of their carrier frequencies, baud rates, etc… Cyclostationary statistics such as the cyclic autocorrelation or the spectral correlation density are some of the statistical terms used to analyse cyclostationarity of signals. These statistical terms have shown to exhibit some features which are very attractive in system identification, such as immunity to coloured noise, robustness in the presence of multiuser interference and other features related to identifiability of nonminimum phase channels. This paper presents a review of the most significant techniques that combine cyclostationary statistics in the time-domain and subspace fitting in the context of multi-antenna system identification. Recent developments and proposals for improvement in direction-of-arrival (DOA) estimation and blind multi-channel estimation and equalisation will be presented.