Learning Vector Quantization-Aided Detection for MIMO Systems

Abstract

In this letter, the learning vector quantization (LVQ) from machine learning (ML) is adopted into the large-scale multiple-input multiple-output (MIMO) detection to improve the detection performance. Inspired by the decision region from lattice decoding, the random Gaussian noises are applied in the proposed learning vector quantization-aided detection (LVQD) algorithm for data generation. Then, based on the classification, supervised learning is activated to update the targeted prototype vector iteratively, so as to a better detection performance. Meanwhile, the decoding radius in lattices is also used to serve as a preprocessing for LVQD, which leads to an efficient detection without performance loss. Finally, simulation results confirm that considerable performance gain can be achieved by the proposed LVQD algorithm, which suits well for suboptimal detection schemes.