Space-Time Algorithms for Mobile Communications
Algorithms are presented for space-time signal processing in mobile communication systems employing base station antenna arrays. Two main issues are concerned. Reception of data transmitted on the uplink (mobile to base station transmission), and estimation of the parameters of the space-time mobile channel.
Receiver algorithms based on iterative least squares are proposed. Channel parameters and signals are iteratively estimated, exploring the finite alphabet property of digitally modulated source signals. The approach is decoupled, treating one signal at a time in a multi user scenario. The basic algorithm, for frequency flat channels, performs spatial diversity combining and is applicable to the antenna array case as well as to CDMA.
The basic algorithm is extended to (1) synchronize a desired signal, and (2) equalize a signal transmitted over a frequency selective channel. Synchronization is accomplished using a scalar synchronization stage, and equalization is performed using an algorithm of RAKE-combiner type. The proposed space-time RAKE-receiver is evaluated using measured GSM array data with excellent results.
To reduce the dimension of the array data prior to reception, a non-parametric beamspace transformation is proposed. The transform is based on a statistical model of the channel, and is designed such that the spatial diversity of the channel is maintained in the lower dimensional beamspace signal.
A weighted least squares approach is proposed for estimation of channel parameters such as directions of arrivals, time delays, and amplitudes. Estimation of directions and delays can be performed either jointly, or in two separate stages. The weighted least squares algorithm is based on knowledge of the source signal, and takes possible modeling errors into account. Parameter estimation with a dually polarized array is given special attention.
Estimation of channel parameter requires knowledge of the model order, that is, the number of propagation paths. A sequential test procedure to detect the number of propagation paths is proposed. The model order is increased until the correct order is found. Several test statistics are proposed, including a parametric test based on full knowledge of the signal model, as well as more robust alternatives for the case of uncertainties in the model. Specifically, robust detection based on the Bootstrap and the Jackknife is considered.
The proposed estimation/detection procedure is applied for evaluation of measured GSM array data. Data collected in one hilly terrain and two different urban macrocell environments are analysed. The results are presented in a variety of ways, for example as space-time power distributions of the mobile channel. The data evaluation also considers estimation for downlink beamforming, as well as polarization parameters.
model order selection