A Study on Replica Generation Using LUT Based on Information Bottleneck for MF-GaBP in Massive MIMO Detection

This paper proposes a symbol replica generation method using look-up table (LUT) based on the information bottleneck (IB) theory in matched filter Gaussian belief propagation (MF-GaBP) for massive multi-input multi-output (MIMO) detection. MF-GaBP serves as an iterative signal detection scheme with low computational complexity by utilizing massive MIMO simplification owing to the law of large numbers. However, when the internal mathematical processes in MF-GaBP are conducted in double precision, a severe processing delay is inevitable. To avoid the impairment, we propose a quantized MF-GaBP detection scheme using predesigned LUTs. The quantization threshold is designed based on the sequential IB (sIB) method for minimizing the mutual information loss. Finally, computer simulations demonstrate that the proposed method significantly reduces the memory occupancy on the basis of table-based processing while suppressing error floor level of the bit error rate (BER) performance.