Subspace-Based Parameter Estimation Problems in Signal Processing

Sammanfattning: The effects of multipath-induced angular spread and non-zero band-width on narrow-band direction-of-arrival (DOA) estimation are investigated. In both cases expressions for the resulting estimation error are developed for the MUSIC, ESPRIT and WSF DOA estimators. The bias expressions are obtained by assuming that the source angular spread and the source bandwidth are small and by then performing a perturbation analysis on the covariance matrix of the array output.Blind (and semi-blind) linear equalizers for direct sequence code division multiple access (DS-CDMA) systems are derived, using multiple antennas in asynchronous intersymbol interference channels. An optimal, in the min-imum mean-square error (MMSE) sense, multi-user DS-CDMA receiver is estimated in a blind (and semi-blind) setting. Simple relations existing between DS-CDMA receivers at the symbol rate and at the chip rate are exploited and deterministic constraints are derived that can be used to determine these receivers in a blind manner, with small sample sizes. The chip rate receiver is shown to be close to a (non-blind) MMSE receiver.A number of subspace-based order estimation methodologies are presented and compared. The methods are all based on matrix rank tests of a Hankel matrix of covariances. Order estimation for both scalar and multi-variable autoregressive moving average (ARMA) processes is emphasized.