Research Article
Highly efficient multiuser detector of M-PSK signals using coordinate-descent phase estimation
@INPROCEEDINGS{10.4108/chinacom.2010.130, author={Yuriy Zakharov and Zhi Quan and George White and Junruo Zhang}, title={Highly efficient multiuser detector of M-PSK signals using coordinate-descent phase estimation}, proceedings={5th International ICST Conference on Communications and Networking in China}, publisher={IEEE}, proceedings_a={CHINACOM}, year={2011}, month={1}, keywords={coordinate descent low complexity M-PSK multiuser detection quadratic optimization semi-definite relaxation}, doi={10.4108/chinacom.2010.130} }
- Yuriy Zakharov
Zhi Quan
George White
Junruo Zhang
Year: 2011
Highly efficient multiuser detector of M-PSK signals using coordinate-descent phase estimation
CHINACOM
ICST
DOI: 10.4108/chinacom.2010.130
Abstract
Joint detection of a large number of M-PSK symbols is important for many applications in communications, in particular, for multiuser detection in CDMA systems. In the past, there was much effort to find a technically simple solution to this problem and yet providing near optimal detection performance. We propose to use a novel iterative technique, the phase descent search (PDS) algorithm, for solving this problem. This technique constrains the solution to have a unit magnitude and it is based on coordinate descent iterations where coordinates are the unknown symbol phases. The PDS algorithm, together with a descent local search (also implemented as a version of the PDS algorithm), is used multiple times with different initializations in a proposed multiple phase detector; the solution with the minimum cost is then chosen as the final solution. Simulation results show that for highly loaded multiuser scenarios the proposed technique has a detection performance that is close to the single-user bound. The results also show that the multiple phase detector allows detection in highly overloaded scenarios and it exhibits near-far resistance. In particular, the detector has a performance that is significantly better, and complexity that is significantly lower, than that of the detector based on semi-definite relaxation.