In this contribution both semi-blind and group blind multiuser detectors designed for uplink MCCDMA systems are investigated in conjunction with different spreading codes. In the investigated scenario Walsh codes are shown to outperform both Gold codes and Zadoff-Chu codes. A short cyclic prefix, rather than differential coding, isused for removing the phase ambiguity imposed by the blind channel estimation scheme, and the Akaike information theoretic criterion advocated by Wax and Kailath is invoked for the estimation of the rank of the semi-blind detectors, which is defined as the number of users activated in the channel. When supporting a total of K = 10 users and exploiting the knowledge of the signature waveforms for the group of K = 7 users communicating within the reference cell considered, while dispensing with any knowledge about the parameters of the remaining K - K = 3 users roaming in the surrounding cells, the system's performance was found to be close to that achieved with the aid of perfect knowledge of all the signature waveforms, the channel coefficients and the total number of users K supported. The achievable performance of this scenario was about 3dB from the single user performance at a BER of 10-3.
展开▼