Soft-decision interference cancellation for AWGN multi-user channels

This work applies the EM algorithm and a more powerful generalized EM algorithm for ML estimation to the multi-user detection problem. A key feature of the resulting receivers is that they use conditional-mean estimates of the users´ bits to reconstruct and cancel multiple-access interference. These soft-decision estimates inherently account for the users SNRs, thereby allowing for effective use of decisions for weaker users in interference cancellation and improving upon the performance of comparable hard-decision-based receivers