A GLR-based Group Detection Strategy for Synchronous CDMA Systems over Frequency-Selective Fading Channels

The Generalized Likelihood Rule (GLR) can be used to demodulate the bits of G out of the K users of a synchronous Code-Division Multiple-Access (CDMA) system without knowledge of the bits, the amplitudes, and the signatures of the interferers. Resorting to the GLR over M + 1 signaling intervals though leads to a detector whose computational complexity is exponential in the number of signaling intervals. Thus, we investigate the suitability of a decision-aided version of the GLR-detector; the proposed receiver is also implemented without knowledge of both the number of interferers, that is determined by resorting to the Akaike Information Criterion (AIC), and the channel impulse response, that is estimated by exploiting the presence of a training sequence (necessary for the initialization of the algorithm). The performance assessment of the proposed detector shows that its loss with respect to a plain Group Detector can be negligible in power-controlled scenarios.