In an urban cellular environment, the data pulses employed in digital communications typically travel through multiple paths from transmitter to receiver due to reflections from buildings and other structures. This multipath propagation causes each pulse to be corrupted by pulses transmitted before and after it; the deleterious effects of this "intersymbol interference" (ISI) necessitate signal processing in the form of an "equalizer". As a further complications, the projected explosion in the number of cellular users will almost certainly require two users in the same of adjacent cells to occupy the same frequency band. This leads to "co-channel interference" that manifests itself as cross-talk amongst users. The use of antenna arrays, which enables discrimination amongst signals arriving from different directions, offers a means of suppressing co-channel interference. This research focuses on the development of spatiotemporal signal processing schemes to effect both equalization and interference cancellation for mobile narrowband digital communications in an urban cellular environment. The equalizer design problem is reformulated as a convex optimization problem based on linear matrix inequalities and solved numerically. Equalization is effected with sample-spaced taps at each antenna, encompassing a time duration roughly equal to the multipath time delay spread. This facilitates better tracking of time-varying multipath channels as compared to conventional equalization schemes employing symbol-spaced taps. Interference cancellation is effected by forming beams with directions of maximum gain steered towards the desired user's multipath, and nulls steered towards strong co-channel interferers. This is achieved by processing a short training signal via a novel extended correlator that highly localizes in time the contribution of the desired user's signal, thereby allowing one to estimate the spatial correlation matrix of the interferers. The equalization an d interference cancellation schemes developed in this research hold the greatest promise in terms of the trade-off amongst convergence rate, computational complexity, robustness to model mismatch, and symbol error performance for Time Division Multiple Access (TDMA) systems where the multipath time delay spread is less than the null-to-null mainlobe of the pulse symbol waveform, as is the case with cellular systems based on the IS-136 TDMA standard.
