Non-linear detectors for multiuser CDMA exploiting non-Gaussianity

Matched filtering can be an effective approach to CDMA detection, even though interfering signals are not truly Gaussian (i.e. there is a model mismatch). Here, two new detectors are derived by assuming very simple non-Gaussian models of the interference probability density functions. We focus on uniform and triangular densities which offer analytic simplicity, and the finite support one would expect from a finite number of power controlled users. The new detectors derived via the maximum likelihood (ML) approach have non-linear structure, and their performance is simulated in actual multiaccess interference (MAI). The performance of this approach turns out to be disappointing, with no gain above ordinary matched filtering unless the interference-to-signal power ratio is extremely high.