Recursive LLR Combining in Iterative Multiuser Decoding of Coded CDMA

Paper in proceeding, 2007

Due to the prohibitive complexity of optimal multiuser detection, suboptimal structures are commonly suggested for use in iterative multiuser decoders. The use of suboptimal components in the iterative structure leads to performance losses in terms of system load, bit error rate (BER), and convergence speed. In this paper, we introduce a simple recursive filtering structure into the iterative multiuser decoder, aimed at improving the performance. Optimal filter coefficients are determined based on the joint likelihood function of the correlated output loglikelihood ratios (LLRs) over iterations. Numerical results show that iterative multiuser systems with recursive LLR filtering and a posteriori feedback at the single-user decoder stage can achieve superior BER performance, produce convergence in fewer iterations, and accommodate an increasing number of active users.