Standard linear diversity combining techniques are not effective in combati
ng fading in the presence of non-Gaussian noise. An adaptive spatial divers
ity receiver is developed for wireless communication channels with slow, fl
at fading and additive non-Gaussian noise. The noise is modeled as a mixtur
e of Gaussian distributions, and the expectation-maximization (EM) algorith
m is used to derive estimates for the model parameters. The transmitted sig
nals are detected using a likelihood ratio test based on the parameter esti
mates. The new adaptive receiver converges rapidly, its bit error rate perf
ormance is very close to optimum when relatively short training sequences a
re used, and it appears to be relatively insensitive to mismatch between th
e noise model and the actual noise distribution. Simulation results are inc
luded that illustrate various aspects of the adaptive receiver performance.