Blind channel identification and space-time equalization for narrow/wide-band digital communications based on parametric modeling of the channel impulse response.

摘要

Transmission of multimedia data is the goal of third generation (3G) cellular radio systems necessitating high-speed data links where the multipath delay spread generally encompasses many symbol intervals. The channel impulse response for each antenna in a prototypical high-speed data link is many symbol intervals in duration and generally sparse. In order to reduce the number of parameters that need to be estimated in order to characterize the channel, researchers have recently proposed a number of schemes for direct estimation of the multipath parameters. However, this approach generally requires a very long observation interval in order to accurately estimate the complex gain of each dominant multipath at each antenna. This dissertation presents an alternative approach based on subbanding and the Cross-Relation Method (CRM) of Xu, Lieu, Tong, and Cliath.; The basic approach is to apply the CRM in each subband, using basis functions derived from the symbol wavering and the bandages filter, to blindly identify the frequency response of each channel in the given subband. In each subband, oversampling of the channel outputs and exploitation of basis function coefficient properties leads to improved performance. However, oversampling can also lead to trivial solutions for which the channel estimates are chosen to be band-reject filters. Two procedures are developed for averting these trivial solutions based on a weighted in-band energy concentration criterion. In addition, a method is presented for proper phasing of each subband channel estimate in order to blindly reconstruct the full channel impulse response associated with each antenna. In the practical commercial case where a small number of training symbols are available, a semi-blind approach is presented that follows the blind procedure by a fractionally-spaced decision feedback equalizer (DFE). The advantage of this semi-blind approach is that the number of taps needed for the DFE need only span the effective duration of the original symbol wavering, which in the wideband case is much less than the number of taps needed to span the delay spread as required in a conventional DFE.; Extensive simulations have been conducted demonstrating accurate identification of multiple channels under multipath conditions, moderate SNR, and with a relatively small number of symbols.