Blind multiuser detection of L out of K active users of an N user synchronous CDMA system

This paper deals with blind multiuser detection in synchronous Gaussian code-division multiple-access systems. In particular, we consider the problem of detecting L out of the K actively transmitting users of an N user system. This problem had been addressed by Varanasi (see IEEE Trans. Inform. Th., vol.41, no.4, p.1083-96, 1995) assuming knowledge of the signatures of all the K active users. The novelty of the present paper is that, while the codes of the L users to be detected are known, those of the remaining K-L active users are not known; moreover, in contrast to previously proposed blind detectors, we make the more reasonable assumption that the codes of the K transmitting users are drawn from a larger set of N codes which are known at the receiver. It is not known which of the K-L out of the N-L possible interfering users are active. The detection strategy we propose ensures higher rates of convergence than previously proposed blind detectors to the corresponding non-adaptive counterparts. In particular, it encompasses as a special case a blind decorrelating detector: the proposed implementation is compared with the blind decorrelator introduced by Wang and Poor (see IEEE Trans. Inform. Th., vol.44, no.2, p.677-90, 1998). The comparison shows that the former is to be preferred to the latter in rapidly varying scenarios.