Design criteria for reduced-complexity joint multiuser detection and group decoding

Decision feedback multiuser detection with trellis code modulation (TCM) has been proposed to increase spectral efficiency of bandlimited multiple-access channel (MAC) [1, 2]. In this paper, we investigate design criteria to achieve good tradeoffs between performance and complexity for joint multiuser detection and group decoding (MD-GD) [2, 3]. Two approaches are considered to improve performance: reduced state algorithms and generalized Viterbi (GV) algorithms. In channels where the main perturbation is multiple-access interference, we show that performance can be further improved increasing the number of states of the channel rather than exploiting the coding information of more users. However, if error propagation is mitigated, additional coding information significantly improves performance. Numerical results show the good performance and low complexity of the MD-GD schemes designed from the new criteria introduced in this paper.