A Low-Complexity Interference-Aware Sphere Detector for Imperfect Channel Estimates

Zhou, G.,Xu, W.,Bauch, G.

DOI: https://doi.org/10.1109/TVT.2015.2453431

2016-01-01

Abstract:In practical applications, an imperfect channel estimate over a noisy channel is inevitable. This paper investigates an interference-aware receiver in which colored estimation noise based maximum likelihood (CEN-ML) detection is employed to mitigate the impact of imperfect channel estimates. It is shown that the CEN-ML detection which treats the channel estimation errors as colored noise has superior performance and is robust to imperfect channel estimation. However, a CEN-ML interference-aware detector usually requires too high computational efforts for multiple-input multiple-output (MIMO) interference-limited systems. To reduce the complexity, this paper proposes new methods for both preprocessing stage and treesearch stage of an interference-aware sphere detector (SD). First, a channel extended matrix is constructed using the information of channel estimation errors in a preprocessing stage. The channel extension model can reduce the complexity without performance loss. Second, two tree pruning criteria are developed in a treesearch stage to make the implementation of the CEN-ML SD feasible. The first criterion is based on a lower bound of the CEN-ML metric and can achieve the same result as the optimal brute force method. The second criterion is derived from an approximated metric and can further reduce the complexity with small performance loss. Simulation results for typical 3GPP LTE/LTE-Advanced scenarios show the new methods provide ability to deal with the channel estimation errors and to achieve a better trade-off between complexity and performance (e.g., in terms of throughput or error rate) in interference-limited scenarios. The proposed low-complexity CEN-ML SD can have a gain of more than 10dB than the other investigated detection for 16-QAM or 64-QAM at a target coded block error rate of 10−2.