A nonlinear filtering approach to estimation and detection in mobile communications

The paper develops an open-loop digital receiver suitable for operation with frequency flat-fading channels disturbed by fast-fading conditions and large carrier phase variations. The receiver consists of a bank of “matched” stochastic nonlinear filters (NLFs) and a maximum a posteriori probability (MAP) decision processor driven by the filters´ innovations processes. This structure is able to perform symbol-by-symbol detection while keeping track of multiplicative distortion (carrier amplitude and phase) within the symbol interval. The NLF propagates probability densities represented as products of Tikhonov and Gaussian functions, whose parameters are computed according to a minimum Kullback (1959) distance criterion, i,e,. we adopt an information preserving strategy. This filter is hangup free and outperforms the extended Kalman-Bucy (1972, 1973) filter (EKBF), especially in terms of phase acquisition. These characteristics explain why the NLF-based receiver may exhibit much lower error probabilities than those obtained with an equivalent EKBF-based receiver. Simulations show that the proposed detection/estimation scheme is robust against mistuning of the receiver parameters, being appropriate for all-digital modems in mobile radio communications